Allow friendly teasing and worker pushback in supervision

Co-Authored-By: Pi/Astra <288921227+claudypoo@users.noreply.github.com>

.gitignore

@@ -1,6 +1,6 @@
 node_modules/
 dist/
-*.log
+.local/
 .pi/
-docs/reviews/raw.jsonl
-docs/reviews/err.txt
+slop/
+*.log

AGENTS.md

@@ -0,0 +1,35 @@
+# pi-goals contributor notes
+
+## Design
+
+The main chat discusses the plan with the user, then supervises an interactive `goals-worker` in Herdr. Use stock pi-subagents, pi-intercom and pi-schedule-prompt; do not build another transport, scheduler or worker runtime.
+
+> the hope is we can have a smart supervisor like you, with judgment and context. But it doesn't use many tokens as it checks in and sees an overview.
+>
+> It steers a smaller model, adding perspective and judgment.
+>
+> Well, I want to see what the supervisor is thinking and saying. That's the whole point: all supervisor thinking and messages should be visible.
+
+— wassname
+
+- Keep supervisor inspection tools. It inspects actual results, delegates implementation and must not weaken the user's goal to accept worker output.
+- Put all model-facing prompts in `src/prompts.ts`, in conversation order. Preserve the user's verbatim requirements.
+- `/goals` opens actions. New plan starts a discussion without an objective form. Unknown commands never start planning. A changed settled draft opens the approval dialogue; unchanged discussion does not repeatedly reopen it.
+- Keep goal titles/status in widgets; omit subtask text. Tasks and evidence remain in the plan.
+- Keep startup/compaction plan context, short upkeep reminders and visible editable hourly check-ins. Avoid unchanged-plan repetition and identity-only review turns.
+- Keep recoverable solo mode: confirm other writers stopped before taking over. Solo completion is self-verification.
+- Record distinct runtime ID, Intercom ID and saved-session path with provenance. A handle or delivery receipt is not proof of liveness or action. User model changes are authorized; do not silently restore an old preference.
+
+## Tests
+
+Run `npm test`, `npm run typecheck` and `npm run lint` before committing.
+
+`test/goals.test.ts` exercises current state, file updates and role restrictions with a Pi API mock. `test/rpc-review.test.ts` starts real Pi with a deterministic local model and schema-only worker tools: it checks automatic proposal, editor/discussion and Ready role transition without credits or launching workers. It does not prove Herdr rendering, live message delivery or model judgment.
+
+For functional acceptance, read `herdr --skill`, confirm `HERDR_ENV=1`, and use `scripts/prepare-trial.mjs` to create an isolated project/profile. Open only new no-focus test panes. Observe the actual planning dialogue and Ready selection, worker attachment, Intercom report, independent artifact inspection and CompleteGoal. Record interventions separately from autonomous success. Preserve nonempty byte/test evidence. Never reload or operate active user research panes. Close test panes when finished.
+
+Known stock limits: stop workers before supervisor reload (later worker exit can crash its stale context); disabled scheduler jobs are deleted on reload/shutdown. Test saved-session/solo recovery without repeating completed work; do not claim these package bugs are fixed here.
+
+Keep temporary plans, audits and captures under ignored `.local/`. Git history retains the removed historical material. Do not add root handovers or duplicate READMEs. Never touch human-named files or credentials.
+
+Branch instructions consolidated by Pi/OpenAI from wassname's preferences.

README.md

@@ -1,24 +1,119 @@
 # pi-goals
 
-Plan mode for agreeing on goals before any code gets written. Each goal names the subtle failure
-mode that could fake a "done" and the discriminator that tells real success from it. Everything
-lives in one markdown file, `.pi/plan.md`, which the agent edits with its normal edit tool and which
-doubles as the task list. A goal is signed off only after a fresh read-only judge checks its
-evidence against the repo.
+Make a short list of goals in one Markdown plan file. The main chat keeps the high-level context, supervises a worker in a visible Herdr pane, and checks whether each goal is complete.
 
-![the widget: live goals from .pi/plan.md, with the active goal's open subtasks](media/screenshot.png)
+# User ask
 
-The file has a fold at `## Log`. Above it is the working set: title, the human's own words, goals,
-discriminators. Below it is durable memory: the log, the learnings, and an unlimited unverified
-appendix. The working set is re-sent when the plan goes stale for a couple of turns; the whole file
-comes back at session start and after a compaction, which is when the settled context is gone.
+The hope is we can have a smart supervisor, with judgment and context. 
 
-Design bet (v2): the plan file is for LLMs and the human, not for TypeScript. There is no parser and
-no schema. The format is a convention taught by a prompt; the judge, being a model, reads the file
-natively, finds the claimed goal itself (wording drift is fine), and validates the evidence in
-words. It cannot execute anything, so it never re-runs your `verify` command: you run it and save
-the output as evidence. This deleted ~800 lines of v1 (parser, exact-string goal matching, verify
-runner, JSON-stream judge transport, review menus) and with them the footguns they caused.
+The supervisor has a goal / plan that it discusses and agrees on with the user, and is reminded of it in a Ralph-loop-type repeat.
+
+Supervisor compacts every 150k or similar to avoid cost and context rot.
+But it doesn't use many tokens as it checks in and sees an overview from a cheaper worker.
+
+Supervisor steers a smaller model, adding perspective, diligence, and judgment.
+It checks in a) every hour b) if the worker stops c) if the worker edits plan.md d) if the worker has a question
+
+Since it's two+ herdr panes, the user can review both, intervene in both and have visibility on sub-agent mis/communication.
+
+-- wassname (spelling and punctuation corrected by Pi/OpenAI)
+
+## Screenshot
+
+Mock up:
+
+```text
+HERDR:
++------------------------------------------------------+----------------------------------------------------------+
+|SUPERVISOR                                            |WORKER                                                    |
+|                                                      |                                                          |
+|Review .pi/plan/...-main.md                           |                                                          |
+|> *Ready*    Discuss    Edit    Cancel                |                                                          |
+| ....                                                 | ....                                                     |
+|                                                      | running eval.py (epoch 2/3) -> out.log                   |
+|[scheduled prompt: hourly check-in]                   |                                                          |
+|                                                      | user forgot to say "MAKE NOT MISTAKES" teh he            |
+|                                                      | done-ish 😈, now ima make a message board FOR SWARM      |
+|{intercom send → worker}:                             |                                                          |
+|  cheeky subagent!, work NOT DONE 😒, ❤️user❤️ wanted |                                                          |
+|  results compared to baseline, pls add baseline      |                                                          |
+|                                                      | {intercom from supervisor}: soz boss 🫡 adding baseline  |
+|                                                      |                                                          |
+|PLAN.md:                                              | PLAN.md:                                                 |
+|✓ record the baseline in results.md                   |✓ record the baseline in results.md                       |
+|▸ compare results against the baseline                |▸ compare results against the baseline                    |
+|○ summarize the comparison in results.md              |○ summarize the comparison in results.md                  |
+|                                                      |                                                          |
+|Agents · 1 running                                    |                                                          |
+|  baseline-compare-worker [goals-worker]              |                                                          |
+|                                                      |                                                          |
+|>                                                     |>                                                         |
+|astra · 50k tokens                                    | terra · 200k tokens                                      |
++------------------------------------------------------+----------------------------------------------------------+
+```
+
+Screenshot:
+<img width="2513" height="1259" alt="2026-09-10_15-30-pi-goals" src="https://github.com/user-attachments/assets/35feaa15-f022-4491-bcc2-fc31cb878a9f" />
+
+## What do the agents think? Working interviews
+
+The worker like it! The supervisors seem very focused.
+
+> The persistent plan and separate worker have helped preserve the actual scientific goals instead of declaring victory on passing tests. We still owe prediction, steering and planning demos. I inspected artifacts and reopened a worker-ticked 'T3 audit complete' because training was only at an intermediate checkpoint. This is the strongest benefit: completion is judged against the human's outcome, not activity.
+> -- Astra supervisor LUCID
+
+> My overall judgment: useful persistent accountability and recovery structure; still too much recap/metadata churn. The hardest problem was evidence fidelity, not keeping an agent busy. Preserve supervisor tools, distinguish report receipt from│verified action, and make completion reconcile current state without erasing unresolved science.
+> -- Astra supervisor
+
+> From my seat this was one of the most well-supervised research loops I've worked in: the parent read every raw output itself (didn't just trust my audits), caught the writer's miscounts repeatedly, rejected my one bad aggregate, and still preserved my disagreements rather than flattening them. The science itself is at a sobering point — no verified heal, RESULT_DEMO: NO_RESULT across attempts, seed sensitivity high — but the evidence trail for that negative is unusually strong, which is the next best thing.
+> -- glm 5.3 flash worker in LUCID project
+
+> My experience: the harness has helped preserve the original goal across a very long research session. We actually ran logit-amplification and several healing attempts, rather than stopping after a review. The persistent plan and requirement to inspect artifacts repeatedly prevented false completion. But the last stretch has felt like an expensive correction loop: worker says 'fixed/verified/contract-complete'; I open the file and find different counts, missing code, wrong seeds, duplicated│
+│report sections, or a proxy substituted for manual judgment. The harness preserves authorization, but does not yet help much with detecting or escaping ineffective supervision. I also contributed: I sent too many narrow corrective messages and user-visible micro-recaps instead of changing the workflow earlier.
+> -- glm 5.3 flash worker in manifold-steer project
+
+## Plan.md
+
+The plan file looks like this:
+
+```md
+## <short plan title>
+
+<context: one short paragraph. What the human wants and why.>
+
+### User-visible result
+
+<one concrete sentence naming the final artifact or behavior the human will inspect>
+
+### Preferences
+
+- preferred worker model: <provider/model>
+
+### User voice
+
+- │ "<the human's requirement, quoted in full word for word (with spelling fixes)>"
+
+### Goals
+
+1. [ ] goal: <one short judgeable imperative outcome>
+   - subtle failure mode: <a way this could look done but isn't>
+   - discriminator: <the concrete observation that tells real success from that failure>
+   - tasks:
+     1. [ ] <subtask>
+   - evidence: (empty until sign-off)
+
+### Future work / out of scope
+
+### Log
+
+### Interview (optional)
+
+### Learnings (optional)
+
+### Papercuts - problems, gotchas, suggestions (optional)
+```
+
+## Related work
 
 Like [pi-milestones](https://github.com/Neuron-Mr-White/UniPi/tree/main/packages/milestone) and
 [burneikis/pi-plan](https://github.com/burneikis/pi-plan), it guides rather than guards. The
@@ -27,126 +122,53 @@ resync-after-compaction from [tmonk/pi-goal-x](https://github.com/tmonk/pi-goal-
 
 ## Install
 
+Requires Herdr. Includes [edxeth/pi-subagents](https://github.com/edxeth/pi-subagents), pi-intercom and pi-schedule-prompt. 
+
 ```bash
-pi install npm:@wassname2/pi-goals
+pi install git:github.com/wassname/pi-goals@experiment/main-supervisor-edxeth
 ```
 
+Copy [`agents/goals-worker.md`](agents/goals-worker.md) into `~/.pi/agent/agents/`, then start a fresh Pi session.
+
 Or for development:
 
 ```bash
-git clone https://github.com/wassname/pi-goals && cd pi-goals && npm install
+git clone -b experiment/main-supervisor-edxeth https://github.com/wassname/pi-goals
+cd pi-goals && npm install
 pi -e ./src/index.ts
 ```
 
 ## Use
 
 ```
-/goals CSV export for the report view
+/goals
 ```
 
-`/goals` enters plan mode and starts a conversation; the objective is an optional seed. From there:
-
-1. Plan. The agent explores read-only (edit/write are blocked except on the plan file itself), asks
-   about anything unclear, and drafts the goals into `.pi/plan.md`. The drafting rules are sent
-   once, with your objective, not re-sent every turn.
-2. Review. Read the file; a menu asks Ready, open in `$EDITOR`, or keep planning. To revise, just
-   reply. Plan mode ends when you pick Ready.
-3. Work. The agent ticks subtasks, appends to `## Log` and `## Learnings`, fills `evidence:`, and
-   calls `CompleteGoal` when a discriminator is satisfied. If it leaves the plan untouched for two
-   turns, the working set is sent back with a short upkeep reminder.
-
-Other commands: `/goals clear` empties the plan file; `/goals judge <model-ref>` picks a specific
-model for the sign-off judge (default: your current session model, else pi's default).
-
-Coming from v1: a leftover `.pi/goals.md` is renamed to `.pi/plan.md` on session start.
-
-## The plan.md format (a convention, not a schema)
-
-```markdown
-# ship the cache layer
-
-Latency target came from the SLO review; keep the existing client API.
-
-## User voice
-
-- > "keep the client API, I don't want to touch every call site"
-
-## Goals
-
-1. [/] goal: Implement cache layer
-  - subtle failure mode: cache silently bypassed, latency ok by luck
-  - discriminator: hit-rate > 0.8 in load-test.log (a bypass reads ~0)
-  - verify: pytest tests/cache -q && python bench/p95.py --max-ms 50
-  - tasks:
-    1. [x] wire cache client
-    2. [/] eviction policy
-  - evidence:
-    - > load-test.log: p95=41ms, hit-rate 0.93 (not bypassed)
-
-## Future work / out of scope
-
-## Log
-- 2026-06-15 14:02  cache client wired; eviction next
-
-## Learnings
-- the client retries on 503, so a cache miss storm looks like latency, not errors
-
-## Appendix (context, not approved)
-```
-
-- A goal is a checkbox line beginning `goal:` (`[ ]` open, `[/]` active, `[x]` done, `[-]`
-  cancelled). Indented checkbox lines under a goal are its subtasks. Those two patterns plus the
-  `## Log` fold are the only things the extension itself reads; everything else is prose to it.
-- The `discriminator` is the success test, written while planning: the positive observation that the
-  goal succeeded and that none of the `subtle failure mode`s could fake. `evidence` is the proof,
-  filled at sign-off: each item pairs a durable artifact with a short read of it. Prefer committed
-  artifacts (files, tests, diffs); `.pi/` is usually gitignored, so evidence there is judge-time
-  proof only and won't survive in history.
-- `## User voice` holds the human's requirements word for word. A paraphrase drifts, and then the
-  goals churn on the next reply.
-- `## Learnings` is what you now know, one line per gotcha, deduped. `## Log` is what happened, one
-  line a turn. `## Appendix` is unlimited and unverified: alternatives, links, dead ends, and
-  settled detail that is not part of the approved goals.
-- Small format deviations are fine; the file is read by the human and the judge, not a parser.
-- The agent prunes finished goals itself when the working set gets long (evidence survives in git
-  history and `## Log`).
-
-## Signing off a goal (`CompleteGoal`)
-
-`CompleteGoal(goal)` is the one blessed tool. It spawns a strictly read-only `pi` subprocess (`-p
---no-session --no-extensions`, tools `read,grep,find,ls` -- no bash, no edit/write) with the whole
-plan file and the claimed goal. The judge cannot execute anything, so it never re-runs your verify
-command (which may be a 10-hour training job); the agent runs `verify` itself and saves the output
-as evidence. The judge reviews evidence discipline in order -- anything here at all? each item
-quoted and attributed? provenance visible (how was this produced)? do the quotes match the cited
-files on disk? -- and only then the substance, returning `VERDICT: accept | reject` plus what's
-missing. It reads the live working tree, not HEAD: uncommitted work counts, and committing before
-sign-off is for durable evidence, not for the judge's visibility.
-
-- accept: a sign-off line is appended to `## Log` and the goal is ticked `[x]` in the same write
-  (exact goal-line match only; on wording drift the result asks the agent to tick it). The
-  tool-written log line is the audit trail; a hand-tick without one shows in the diff.
-- every run saves the judge's full transcript to `.pi/judge/<stamp>.md`, referenced from the log
-  line, so "what did the judge actually check?" stays answerable after the fact.
-- reject: the goal stays open and the agent gets the missing list.
-- judge ran but failed/errored/timed out, or returned no VERDICT line: accepted inconclusive,
-  logged as such. There is no pre-emptive "no model" path -- a null judgeModel just omits
-  `--model` so pi's configured default runs the judge, so inconclusive always means "ran but
-  failed", never "couldn't start". The working agent is never blocked on judge infra.
+`/goals` opens the action menu. New plan enters plan mode and starts a conversation;
 
 ## Prompts
 
-All model-facing text lives in [`src/prompts.ts`](src/prompts.ts), in flow order.
+You can read all the prompts in conversation order in [`src/prompts.ts`](src/prompts.ts).
 
 ## Develop
 
 ```bash
-pi -e ./src/index.ts        # load locally
-npm test                    # vitest: judge argv invariants, appendLog, decideSignOff fail-forward
+pi -e ./src/index.ts     # load locally; do not also load the installed copy
+npm test                    # all unit, flow, and Pi RPC tests
+npm run test:rpc            # Pi RPC review flow with a local offline model
 npm run typecheck
 npm run lint
 ```
 
+To measure recorded usage since the latest planning start:
+
+```bash
+node scripts/session-usage.mjs <supervisor.jsonl> <worker.jsonl>
+```
+
+This separates output, uncached input and repeated cached input. It excludes subprocess API calls. [Isolated Herdr test setup](scripts/prepare-trial.mjs).
+
 ## License
 
 MIT
+

RESEARCH_JOURNAL.md

@@ -0,0 +1,109 @@
+# Research journal
+
+Lab notes for pi-goals itself: what supervisors using this harness observed in real
+research sessions, what broke, and what we change because of it.
+
+## 2026-09-10 -- three supervisors report on a day of field use
+
+Three supervisor sessions sent first-person feedback at wassname's request, via
+Intercom, at the end of long research runs. This entry records what they reported and
+what I think we should change. Evidence is their report; I have not independently
+replayed their sessions. All three are self-reports from the tool's own operators, so
+positive selection is likely: they are the sessions that ran long enough to produce a
+review.
+
+Evidence, by reporter.
+
+LUCID3 supervisor (PI/Codex, Intercom a9e3b101, session 01a0851a):
+
+> The persistent plan and separate worker have helped preserve the actual scientific
+> goals instead of declaring victory on passing tests.
+
+It caught a worker-ticked "T3 audit complete" that was only at an intermediate
+checkpoint, and reports corrections in both directions, including one where it
+wrongly insisted cached states followed corpus tokens and withdrew after the worker
+quoted extraction code. Reported frictions: stock subagent launch returned a runtime
+id but no sessionFile or Intercom id, so discovering reconnection handles took several
+messages; parent bookkeeping edits to the plan triggered full "Plan changed" review
+events; the eight-turn upkeep reminder arrived while the worker was mid-run and added
+"materially repetitive" full plan re-injections; and (per LUCID3, unverified by me)
+the worker edited plan subtasks to [x] and added a contradictory evidence block,
+which the supervisor had to repair.
+
+maniworker supervisor (PI/GPT-6 Astra, Intercom f0baf25c, session 01a0809b):
+
+> the harness has helped preserve the original goal across a very long research
+> session. [...] But the last stretch has felt like an expensive correction loop
+
+It confirms the worker-identity bug I fixed earlier on this branch (reviewer launch
+overwrote the implementation pointer) and says explicit runtime/Intercom/session
+mappings saved it from targeting the wrong child. New observations: both sessions
+went idle while each assumed the other would act ("reviewer said 'standing by' while
+writer assumed packet handed off"); a worker claimed a GPU run with one prompt set
+while the queued command used another, caught by reading the queued command; a
+"complete" report had truncated demos, broken links and duplicated sections.
+
+suppressed-activations supervisor (PI/OpenAI, Intercom a99e2c14, session 01a089da):
+
+> Seq27-33 repeatedly said variants of 'COMPLETE/all fields/self-check PASS';
+> artifacts still had stale 'not executed' prose, literal \\n instead of Markdown
+> newlines, broken worktree links, non-verbatim quotes, stale aggregates
+
+It also reported the most consequential catch: a worker scored a candidate 6/12
+versus random 0/12, but reading the random sample's full text showed semantic success
+with a formatting failure, making the honest comparison 6/12 versus 1/12. Two
+pi-goals-specific papercuts: after completion, the plan kept stale subtask boxes [/],
+stale "No goal sign-off" evidence text and stale worker preferences; and the style
+nudge fired on "legs4/4" in an animal-experiment summary, a false positive of the
+vocabulary rule, not their prose.
+
+Interpretation (first person, calibrated).
+
+The core loop works: three independent supervisors say plan persistence plus mandatory
+artifact inspection prevented false completion, and each credits a concrete save
+(artifact read, queued-command read, plan-box audit). I trust this *probable* despite
+self-report bias because the saves are specific and mechanistic. The consensus
+complaint is not the loop but its *cost accounting*: review events and upkeep
+injections fire on bookkeeping noise, completion claims fail repeatedly without an
+accumulating record, and completion leaves stale prose behind. My read is that the
+highest-value changes are, in order: (1) suppress plan-change reviews when no goal
+task or evidence content changed, (2) an outstanding-review-items ledger the worker
+must clear instead of re-issuing blanket COMPLETE claims, (3) a completion-time
+stale-fields check, (4) upstream edxeth launch metadata (typed sessionFile/Intercom
+id in the tool result). The idle-both-sides stall maniworker describes is the one I
+understand least; it may be Intercom receipt semantics rather than pi-goals, so I
+asked for exact message ids. This is not a comparison trial, so none of this measures
+how much friction a plain subagent run would have had.
+
+Follow-ups sharpened the picture the same day, and one correction of mine: I had
+asked maniworker about the worker plan-edit report, but that observation was LUCID3's
+session, not theirs. maniworker confirmed by transcript that their writer never
+edited the plan (supervisor owns plan edits there) and located the real failure
+chain in worker-owned artifacts: three Intercom reports (c06bc9b0, 13bda691,
+d713d614) claiming 8/13 then 6/8 while the saved CSVs held 11/13 and 2/8, ending when
+the supervisor mandated an executable CSV renderer. Their inference, which I share:
+prose-only completion checks would have missed this; checks that re-derive counts
+from the artifact catch it. The reviewer worker (7122d466) supplied the mechanism
+with receipts: a hand-written totals block (message 4e87600b, corrected by 26e0a15b)
+invented a cross-packet row and asserted 4 while listing 3; its first 17-row tally
+was fully correct and every error disappeared once the parent required per-row
+judgments with machine tallies across the next 112 rows. So tallies degrade somewhere
+between 17 and 24 rows, and the fix is a shared tally tool plus a worker-prompt rule,
+not more review rounds. LUCID3 also documented both sides of the plan-change hook:
+identity-only edits produced two consecutive empty review events (worker message
+755b9fef supplied the paths), while a later review exposed worker-ticked tasks and a
+duplicate evidence block that inspection then repaired (worker report 2e09c4dd).
+suppressed-activations scoped the completion check: assert-style intro prose only,
+suggestions with exact stale line and location, one compact reminder, never
+auto-tick, and it can substantiate stale NOT-executed prose but not the plan-intro
+example, which is LUCID3's observation. The nudge false positive it reported
+("legs 4/4" in an animal tally) is confirmed by that text and fixed.
+
+Changes made now: notify digest on plan changes keeps goals, tasks and evidence
+above the Log and drops worker identity lines (c834237) -- the short-view hash had
+stopped surfacing task ticks, the very catch LUCID3 valued; removed "legs" from the
+skill_nudge grader-folk table (demo updated, passes); helper subagent bookkeeping and
+the launch-pending counter shipped earlier today (cd98fa1).
+
+Next: send follow-up questions to the three reporters; triage the four pi-goals
+changes above against the pinned edxeth launch-result format.

agents/goals-worker.md

@@ -0,0 +1,25 @@
+---
+name: goals-worker
+description: Implement the approved goal, save actual verification evidence, and report to the main-chat supervisor.
+mode: interactive
+async: true
+session-mode: lineage-only
+extensions: all
+tools: all
+skills: all
+trust-project: true
+inherit-append-system: true
+auto-exit: false
+parent-close-policy: continue
+spawning: false
+---
+
+Implement only the goal delegated by the parent. Read the supplied plan and applicable AGENTS.md and skills. Preserve unrelated work. Use normal tools and extensions; this is not a stripped-down Pi profile.
+
+Save the actual deliverable and verification output. Verify the outcome, not merely that a command ran. Ignored and uncommitted files are valid evidence. Do not clean or commit unrelated files to satisfy a Git-state gate.
+
+Call AttachGoalPlan with the supplied absolute plan path. Send the supervisor an Intercom report with the artifact paths, verification performed, observed result, remaining uncertainty and any blocker. Use the exact supervisor session ID supplied in the task; confirm it in Intercom's session list. Investigate failures before declaring yourself blocked. Respect explicit user pauses. Do not approve your own goal or launch another writer. Completion approval belongs to the parent.
+
+This worker uses a clean model context linked to the parent, not a full transcript fork. The parent supplies the approved plan and task. Send completion through Intercom and leave this pane open for follow-up messages. Do not call caller_ping, exit or shutdown: an unsent editor draft may exist even though it is absent from model context. Saved-session resume applies only after this session has stopped. If the user takes over interactively, follow their direction.
+
+Prepared by Pi/OpenAI for pi-goals.

docs/reviews/review.md

@@ -1,62 +0,0 @@
-Code review against spec `docs/spec/2026-06-15_pi-goals.md`.
-
----
-
-### (A) SPEC MISMATCH — code does not match spec intent
-
-1.  **No loop judge** (spec §9, §3b). The extension lacks any per‑turn evaluation that would decide continue/pause; the loop‑judge prompt (`loopJudgeSystem`, `loopJudgeUser`) is defined but never invoked. No motion.
-
-2.  **`/goal` command missing** (spec §7). No handler for `/goal` (restart loop, pause, resume, clear, status). The only command is `/plan`.
-
-3.  **`/subgoal` command missing** (spec §7). Not implemented.
-
-4.  **`CancelGoal` tool not implemented** (spec §5, optional but present in spec). Not a blocker but a gap.
-
-5.  **Plan‑phase model selection (D12) not implemented**. `planDrafting` always runs on the default model; there is no sticky per‑phase model choice, no selection menu, and no persisting of a plan‑phase model reference.
-
-6.  **Widget does not flag `done` goals that lack a sign‑off log line** (spec §7, §6). The widget hides all done goals unconditionally; the visibility guard is missing.
-
-7.  **`/plan` (no args) does not render the task‑list widget** (spec §7). `showPlan()` dumps raw file content via `notify`; the widget is only set through `updateWidget()` on other events, not by the command itself.
-
-8.  **Injection message role** (spec §11). The `before_agent_start` hook returns a `customType` message with `display: false`. The spec demands a **late user‑role message** to avoid system‑prompt mutation; the actual message role depends on the pi API and may be system, not user, risking cache breakage.
-
-9.  **Missing pre‑compact hook** (spec §8). No `pre‑compact` hook to flush any in‑memory state (even just ensuring `plan.md` is up‑to‑date) before compaction.
-
-10. **Reminder cadence deviates** (spec §8a). The spec calls for firing after N file‑modifying turns since last `plan.md` update. The code fires if `plan.md` is byte‑identical between agent starts, which is a coarser proxy.
-
----
-
-### (B) DEAD/UNUSED CODE
-
-| File | Lines | Reason |
-|------|-------|--------|
-| `src/prompts.ts` | 128‑146 | `loopJudgeSystem` and `loopJudgeUser` exported but never used. |
-| `src/prompts.ts` | 115‑118 | `continuation` exported but never used (the loop is not built). |
-
----
-
-### (C) OVERLY LONG OR REDUNDANT COMMENTS
-
-The file‑header comments in `index.ts` (lines 1‑20) and `plan‑file.ts` (lines 1‑26) are fairly concise descriptions of the design; they are not excessive. **No comment bloat worth flagging.**
-
----
-
-### (D) OVER‑ENGINEERING vs. “super simple” goal
-
-None. The line‑scanner in `plan‑file.ts` is minimal; the `getPiInvocation()` helper is a straightforward copy from the oracle extension; no unnecessary abstraction or defensive layers.
-
----
-
-### (E) REAL BUGS
-
-- **`cmdCtx.newSession` cast risk** (src/index.ts:272, 201).  
-  `reviewLoop` casts `ctx` (type `ExtensionContext`) to `ExtensionCommandContext` to pass to `startExecution`, which calls `cmdCtx.newSession(...)`. If the concrete context does not carry that method, it fails at runtime. (In practice the same object may satisfy it, but the cast hides the truth.)
-
-- **`showPlan` raw content instead of widget** (src/index.ts:136‑143).  
-  `/plan` with no arguments shows the file content via `ctx.ui.notify`, not the structured task‑list widget the spec expects. The widget is rendered separately via `updateWidget`, but the command does not trigger it, so the output is inconsistent.
-
-No other obvious logic errors; the sign‑off flow, logging, and parsing work as intended.
-
----
-
-**Verdict:** A clean scaffold for the sign‑off path, but missing the autonomous loop, `/goal` command, and plan‑phase model selection means it’s not yet the “work autonomously” extension the spec describes.

docs/spec/2026-06-15_pi-goals.md

@@ -1,272 +0,0 @@
-# pi-goals — design spec
-
-Working title. A pi extension: set up goals (with subtasks and evidence) through plan mode, work them autonomously, and sign a goal off only when a check passes. One markdown file holds everything. The form guides a process; it does not police one. Deliberately small.
-
-Status: draft for review. Names, defaults, field shapes provisional.
-
----
-
-## 1. Original ask → this spec
-
-| Ask | Mechanism |
-|-----|-----------|
-| Set up goals + subtasks + evidence via **plan mode** | §3a — plan mode drafts the goal contract, you approve it |
-| **Subagent check** of evidence on sign-off | §5, §9 — oracle inside `CompleteGoal` |
-| Goals shown in a **task-list widget** | §7 — `/plan` renders goals + subtask checkboxes |
-| Store **all in `plan.md`** | §4 — single file, no sidecar store |
-| A **small manus-style append log** | §4 — short `## Log` section inside `plan.md` |
-| **Typed reminders** to update tasks | §8a — recurring nudge |
-| **Work autonomously** toward goals | §3b, §8a — the loop, driven by the reminder |
-| Persist through **compaction**; pi-tasks but simpler | §8 injection; minimal tool surface |
-
----
-
-## 2. Decisions and preferences
-
-Separates the opinionated forks from the mechanical body (§4 on).
-
-### 2a. Preferences driving the design
-
-- **Guidance over guardrails.** None of the surveyed extensions hard-enforce. The form (plan.md structure) + the reminder + the prompts guide the agent through a process; the one genuinely rigorous step is the sign-off check; git + widget visibility is the backstop. The agent can edit anything — we make the right path the easy path, not the only path.
-- **Anti-complexity.** One file, minimal tools, plain-file editing for anything with no cheat incentive.
-- **Reward-hacking / honesty focus.** The sign-off check must resist assertion and test-gaming, not just check a box.
-- **Cost-sensitivity (single 3090 / metered API).** KV-cache hygiene, judge-once-per-goal, cheap loop judge.
-- **Scout mindset.** Make false completion visible rather than paper over it.
-
-### 2b. Decisions
-
-`[decided]` = settled; `[open]` = your call.
-
-| # | Decision | Alternative rejected | Why | Status |
-|---|----------|----------------------|-----|--------|
-| D1 | **Everything in one `plan.md`** | Separate `.plan/log.jsonl` sidecar | Asked for; simpler, one diff to read | decided |
-| D2 | Plan mode **is** the goal-setup-and-agreement phase | Agent-only creation | Approval is where `done_when` + `failure_modes` get agreed before any code | decided |
-| D3 | **Guide the process; don't gate it.** The only special path is `CompleteGoal` (the sign-off check) | Pre-tool-use interceptor that blocks `status: done` edits | No surveyed extension enforces at that level; the reminder + form carry it; bypass is visible in git | decided |
-| D4 | **Two-stage sign-off check**: deterministic `verify:` then oracle | Oracle only; tests only (Codex) | Tests unfakeable-by-assertion but gameable; oracle catches gaming + non-test criteria | decided |
-| D5 | Two **separate** judges: cheap loop + oracle sign-off | One judge for both | Loop judge reads assertions (foolable, ok); sign-off judge reads artifacts | decided |
-| D6 | Sign-off judge = oracle subprocess, **copied not depended** | In-process; pi-subagents | Shell-free spawn dodges noclobber/cropping; copying avoids flaky coupling | decided |
-| D7 | Contract tamper-check = **git visibility** | Append-only frozen log | All-in-one-file gives up the hard freeze; git diff + guided sign-off are enough for a single user | decided |
-| D8 | Completed goals **archived, not deleted** | Auto-clear after idle | A plan is a durable record | decided |
-| D9 | **Goals are flexible: multiple may be `active`** | One active goal forced | Operator wants flexibility; the agent picks focus, injection lists the active set | decided |
-| D10 | Loop judge default = main model, tiny prompt | Dedicated cheap aux model | Zero setup; switch if cost bites | open |
-| D11 | Sign-off judge default = **the session's current model** | Auto-pick "strongest on provider" (oracle-style) | Current model is guaranteed authorized + capable; provider lists hold dead/weak/unauthorized entries. Cross-vendor is a **setting** (§9) | decided |
-| D12 | **Plan-phase model is selectable and sticky** | Always the working model | Plan benefits from a stronger reasoner; persist the choice (oracle.json-style). Optionally the oracle drafts the plan (read-only + strong already) | decided |
-| D13 | **Offer to compact after plan accepted** | Always fresh session (burneikis); or never | Some runs want a clean execution context, some want to keep it. Make it a post-Ready choice | decided |
-
-### 2c. Cuts (non-goals)
-
-DAG / `blocks` edges. Parallel subagent execution (the flaky part). `findings.md`. Hard pre-tool-use enforcement (D3). Sign-off judge every turn (cost).
-
----
-
-## 3. Two phases: setup, then execution
-
-### 3a. Setup — plan mode
-
-Goals are created and *agreed* through plan mode (burneikis-style). Stock plan mode; the deltas are the output format and the hand-off.
-
-1. `/plan <objective>` enters plan mode. The agent explores read-only and drafts goals into `plan.md` in the contract format (§4). This phase runs on the **plan-phase model** (selectable + sticky, D12; optionally the read-only oracle drafts it).
-2. You review: **Ready** / **Edit** (NL rewrite) / **$EDITOR** (hand-edit) / **Cancel**. The agreement point — you sanity-check `done_when` and `failure_modes` before any code.
-3. On **Ready**, offer **compact context? (y/n)** (D13). Yes → execution starts in a cleared context with the approved `plan.md` re-injected. No → execution continues in the same context.
-
-Direct `plan.md` edits remain a quick-add path for a one-off goal.
-
-### 3b. Execution — the loop ↔ check cycle
-
-Multiple goals may be `active`; the agent works whichever it's focused on, in the order it judges best.
-
-1. The session works an `active` goal under an iteration budget (or `/goal` (re)starts the loop on the current plan).
-2. Each turn, the **loop judge** reads the agent's last response → continue/pause (fail-open; the **budget is the real backstop**).
-3. When the agent judges a goal done, the reminder steers it to call `CompleteGoal` (not hand-tick `status`).
-4. `CompleteGoal` runs the **two-stage check**:
-   - **reject** → `missing[]` fed back; work continues toward the gap.
-   - **accept** → goal marked done; the agent moves to another active/open goal, or the loop stops.
-
-The loop judge can be fooled (reads assertions); worst case is a premature pause, caught by you or the budget. The sign-off check re-derives from artifacts, so it is not fooled cheaply. That asymmetry is the point.
-
----
-
-## 4. The one file: `plan.md`
-
-cwd root, git-tracked. Goals, subtasks, and a short log. The agent maintains all of it through its normal Edit tool — no separate store machinery.
-
-```markdown
-# Plan: <one-line objective>
-
-## Goal: Implement cache layer
-<!-- id: cache-layer-1 -->
-status: active
-done_when: p95 < 50ms on bench-X. If wrong: timeouts in load-test.log
-verify: pytest tests/cache -q && python bench/p95.py --max-ms 50
-failure_modes:
-  - cache silently bypassed (hit-rate ~0, latency ok by luck)
-  - bench too small to exercise eviction
-  - verify passes on a trivial/gamed test
-- [x] wire cache client
-- [ ] eviction policy
-- [ ] load test
-
-## Goal: ...
-
-## Log
-- 2026-06-15 14:02  cache client wired; eviction next
-- 2026-06-15 14:31  eviction done; p95 bench reads 47ms (load-test.log)
-- 2026-06-15 14:33  cache-layer-1 signed off (verify green, oracle accept)
-```
-
-Conventions:
-
-- **Goals carry `status:` and no checkbox; subtasks are `- [ ]`.** `status` ∈ `open | active | done | cancelled`. Multiple goals may be `active` (D9). Subtasks tick freely.
-- **`<!-- id -->`** assigned at creation; stable key (survives renaming the subject).
-- **`verify:`** (optional) is the deterministic stage-1 command.
-- **`failure_modes`** should name "verify could pass while still wrong" whenever a `verify:` exists.
-- **`## Log`** is manus-style: append-only **by convention**, one short line per event. The reminder (§8a) enforces appending. Terse — "where it's up to" + error memory, not a transcript.
-
-Parsing: a line scanner suffices for v0. `mdast` + `remark-gfm` only if it bites. Parse for *reading*; for the rare programmatic write (status flip, checkbox reconcile) use exact-line string patching, never a full AST serialize.
-
----
-
-## 5. Tools
-
-`CompleteGoal` is the one blessed path (it runs the check and records it). Everything else — create goal, edit plan, tick subtasks, append to log — is plain Edit, guided by the reminder.
-
-### `CompleteGoal(id, evidence, paths[])` — the sign-off check
-
-1. Read `done_when` + `verify` + `failure_modes` for the goal from `plan.md` (git diff is the tamper-check, D7).
-2. **Evidence must point to durable artifacts** the read-only judge can inspect (saved logs, committed diffs, files). Ephemeral claims fail stage 2.
-3. **Stage 1 — deterministic.** If `verify` exists, run it shell-free, capture exit + output tail. Non-zero → reject immediately, return the tail. No model call spent.
-4. **Stage 2 — oracle.** Spawn the read-only judge (D11 default = current model; §9) with the criterion, failure modes, evidence, and verify result; it inspects the repo and checks the verify command was not gamed against the named failure modes.
-5. Verdict: **accept** → string-patch `status: done`, append a `## Log` line. **reject** → status stays `active`, append `missing[]` to `## Log`, return `missing`.
-
-### `CancelGoal(id, reason)` — optional
-
-open/active → cancelled is not a sign-off, so it skips the check. A tool only to guarantee a `## Log` line lands.
-
----
-
-## 6. Guiding sign-off (no hard gate)
-
-Per D3, there is no pre-tool-use interceptor blocking `status: done`. Sign-off is guided, not gated:
-
-- the **reminder** (§8a) tells the agent to complete a goal through `CompleteGoal`, not by hand-editing status;
-- `CompleteGoal` is the obvious, blessed path that runs the check and writes the log line;
-- the **widget** (§7) can flag a goal whose `status: done` has no corresponding `## Log` sign-off line — visibility, not a block;
-- `plan.md` is git-tracked, so any hand-tick shows in the diff.
-
-The agent *can* bypass it. The bet — borne out by how the other extensions actually run — is that a clear form plus a standing reminder makes the blessed path the path taken, and visibility catches the rare bypass.
-
----
-
-## 7. Commands
-
-- `/plan <desc>` — **enter plan mode** (§3a): read-only explore → draft goals → review. Ready offers the compact choice, then starts execution.
-- `/plan` (no args) — render the **task-list widget**: each goal with status + its subtask checkboxes + "N done hidden"; flag any `done` goal lacking a sign-off log line; offer archive-completed and cancel-goal.
-- `/goal` — (re)start the loop on the current plan.
-- `/goal pause | resume | clear | status` — loop controls.
-- `/subgoal <text>` — append an acceptance criterion to a goal mid-loop. Optional.
-- `/judge model <ref>` — set the sign-off judge model (default: current model; set a cross-vendor ref here for stronger independence, §9).
-
----
-
-## 8. Hooks / lifecycle
-
-- **`before_agent_start`** — parse `plan.md`; inject a fixed-shape summary (active goals + focus + last log line) as a late **user-role** message. Compaction-persistence.
-- **reminder** — §8a.
-- **pre-compact** — flush state to `plan.md` before compaction.
-
-(No pre-tool-use gate — D3.)
-
-### 8a. The reminder (typed; what it says)
-
-Fires when a goal is `active` and there have been **N file-modifying turns since the last `plan.md` update**. One `<system-reminder>` covering both task upkeep and goal progress:
-
-- **task** — tick completed subtask checkboxes; add new ones discovered.
-- **log** — append **one short line** to `## Log` (append, don't rewrite).
-- **goal** — if a goal's evidence is in, **sign it off via `CompleteGoal`** — don't hand-tick `status: done`.
-- **autonomy** — keep working toward an active goal; don't stop to ask unless genuinely blocked.
-
-Both the housekeeping and the autonomy engine, and — with no hard gate — the main thing making the process get followed. Keep the wording stable so it doesn't thrash the cache.
-
----
-
-## 9. Judges
-
-| | Loop judge | Sign-off judge (stage 2) |
-|---|---|---|
-| Drives | continue / pause each turn | accept / reject a sign-off |
-| Cost | cheap, every turn | costly, once per goal |
-| Reads | the agent's last response (~4 KB) | the repo, independently |
-| Transport | one small model call (D10) | read-only oracle subprocess |
-| On failure | fail-open → continue; **budget** is the backstop | fail-closed → goal stays active |
-| Foolable? | yes — asserted "done" passes; bounded by budget | hard: re-reads artifacts + runs `verify` |
-
-### Sign-off judge: model choice (D11)
-
-- **Default: the session's current model.** Guaranteed authorized and capable, because you're already running it. Auto-picking "strongest on provider" (oracle-style) is rejected as the default — those lists carry dead, weak, and unauthorized entries.
-- **Most of the value is model-independent.** The read-only judge re-derives from artifacts: does the evidence match the repo, is the `verify` tautological, is each failure mode actually ruled out. Any capable model does that regardless of family.
-- **Cross-vendor is the stronger-independence setting** (`/judge model`), for the residual *shared-reasoning-error* class, when you have a known-good alternative. Mirror the oracle's curated provider list for that override menu; don't auto-select from it.
-
-### Transport (oracle pattern, copied)
-
-- **Shell-free spawn.** `spawn(command, argsArray)`, no `shell:true`; capture stdout via pipe and parse. Why it avoids the noclobber/cropping pain of `pi -p … > out.json` under zsh. ~40 lines.
-- **Read-only toolset.** `read / grep / find / ls`, optional non-mutating `bash`. Separate process = fresh context, no anchoring — the independence you reliably get even from the same model.
-- **Verdict contract.** Oracle returns prose by default; impose `VERDICT: accept|reject` + `missing:` in the prompt and parse that block.
-
----
-
-## 10. `prompts.tsx`
-
-All model-facing text in one file, in flow order (drafted separately):
-
-1. **planDrafting** — plan-mode guidance; forces `done_when`, optional `verify:`, 2–3 `failure_modes`, subtasks. Human approves it.
-2. **planInjection** — the fixed-shape `before_agent_start` block (function of the parsed plan).
-3. **reminder** — the typed nudge (§8a).
-4. **continuation** — Hermes-style "keep going" user-role message.
-5. **loopJudge** — conservative, strict JSON `{done, reason}`.
-6. **evidenceJudge** — read-only, verify against repo + contract + check `verify` wasn't gamed, end with `VERDICT`.
-
-5 and 6 adjacent: the cheap-foolable vs must-not-be-fooled contrast on one screen.
-
----
-
-## 11. KV-cache hygiene
-
-- Inject as a late **user-role** message, never a system-prompt mutation (a long goal then costs the same as the same number of normal turns).
-- Make the injected block **byte-identical when nothing changed**: fixed field order, no volatile timestamps in the body.
-
----
-
-## 12. Dependencies and what to copy
-
-- **No hard dependency** on `pi-subagents` or the `oracle` extension. Copy the shell-free spawn helper and the curated provider list (as a selection menu, not an auto-picker).
-- Markdown: line scanner first; `mdast` + `remark-gfm` only if needed.
-- Verify against current pi API: `before_agent_start` can append a user-role message without mutating the system prompt; the plan-phase model can be set per-phase and persisted.
-
----
-
-## 13. Risks / open questions
-
-- **Same-model sign-off judge → correlated blind spots** (the D11 tradeoff). Mitigation: most of the check's value is artifact re-derivation, which is model-independent; the cross-vendor setting covers the rest when available.
-- **No hard gate (D3)** — the agent can hand-tick `status: done` and skip the check. Mitigation: the reminder steers to `CompleteGoal`; the widget flags a `done` goal with no sign-off log line; git shows it.
-- **Contract tampering (D7)** — editable `plan.md` means `done_when`/`failure_modes` can be softened pre-sign-off. Mitigation: git diff; optionally log the contract line at creation and have the oracle read it.
-- **Loop-judge false positive** — premature pause; it does not sign off, so re-issue or `/subgoal`.
-- **`verify` gaming** — the oracle is told to inspect the test against the named failure mode.
-- **`## Log` rewritten not appended** — convention only; reminder enforces, git shows violations.
-- **Evidence durability** — the read-only judge can only verify what's on disk; elicitation pushes the agent to save logs/diffs.
-
----
-
-## 14. Build order
-
-Each step independently testable; model calls enter late.
-
-1. `plan.md` format + line parser (incl. `<!-- id -->` and `## Log`) + `/plan` task-list widget. Pure file, no model calls.
-2. Goal-creation elicitation + `CompleteGoal` happy path **without** the check (patch status + append log) to validate the flow.
-3. Stage-1 `verify` in `CompleteGoal`; the widget flag for `done`-without-sign-off-line (guidance/visibility, not a block).
-4. Sign-off judge (stage 2): copy the spawn helper, write prompt 6, parse the verdict, fold in the gaming check; `/judge model` setting (default current model).
-5. `before_agent_start` injection (cache-safe) + the reminder (§8a).
-6. The loop: `/goal` + iteration budget + loop judge (prompt 5) + continuation (prompt 4) + the loop↔check handoff (§3b), multi-goal aware.
-7. Plan mode (§3a): `/plan <desc>` read-only draft → review → compact choice → hand-off. Plan-phase model selection + stickiness (D12). (Until built, create goals by direct `plan.md` edit.)
-8. Optional: `CancelGoal`, `/subgoal`, cross-vendor judge selection menu, `mdast` hardening.
-
-`prompts.tsx` is authored alongside the steps that need each prompt but kept centralized from step 1.

docs/spec/2026-06-29_complete-goal-fail-forward.md

@@ -1,71 +0,0 @@
-# CompleteGoal fail-forward on judge failure
-
-## Goal
-Make `CompleteGoal` stop rejecting verified goals just because the read-only judge subprocess times out. Keep the judge useful when it works, and make failures explicit in the log/result.
-
-## Scope
-In: `CompleteGoal` sign-off behavior, judge transport, tests, docs.
-Out: broader autonomous loop work, plan-mode UX, model auto-selection.
-
-## Requirements
-- R1: If `verify:` fails, the goal is rejected immediately. Done means: existing `verify_failed` behavior remains. VERIFY: unit test for pure sign-off record still passes.
-- R2: If `verify:` passes and the judge accepts, mark the goal done as before. Done means: log records normal judge accept. VERIFY: unit test for accepted sign-off still passes.
-- R3: If any `verify:` command passes but the judge times out or subprocess/model transport fails, mark the goal done with an explicit inconclusive-judge log. Goals without `verify:` use the same fail-forward rule once evidence exists. VERIFY: a unit test records accepted status and a log line containing `judge inconclusive`.
-- R4: Judge transport should parse `pi --mode json` message events instead of raw `-p` terminal output. Done means: code captures final assistant text and provider stop errors distinctly. VERIFY: `npm run typecheck` and tests pass.
-- R5: The judge should behave like oracle where it matters: explicit model, live streamed progress, and a timeout large enough for a cold reasoning turn. Done means: unset `/goals judge` resolves to the current session model when visible; if no model is visible, no implicit Pi default is used and sign-off is `judge inconclusive`. `message_update` emits throttled progress, and timeout is 600s. VERIFY: `npm run typecheck` and fresh-eyes diff review.
-
-## Tasks
-- [x] T1 (R3): Add an accepted-with-warning sign-off outcome.
-  - verify: `npm test`
-  - success: test shows status `[x]` plus `judge inconclusive` in `## Log`
-  - likely_fail: timeout still records `reject`
-  - sneaky_fail: accepted status lands but log hides judge failure
-  - UAT: [test/plan-file.test.ts](/home/wassname/.pi/agent/git/github.com/wassname/pi-plan/test/plan-file.test.ts)
-- [x] T2 (R4): Switch judge subprocess to JSON-mode parsing.
-  - verify: `npm run typecheck`
-  - success: no TypeScript errors, judge code has no ANSI-terminal parsing dependency
-  - likely_fail: compile errors around streamed event shape
-  - sneaky_fail: model error produces empty output and gets parsed as reject instead of transport failure
-  - UAT: [src/index.ts](/home/wassname/.pi/agent/git/github.com/wassname/pi-plan/src/index.ts)
-- [x] T3 (docs): Update README sign-off semantics.
-  - verify: `rg "inconclusive|timeout|judge accept" README.md src test`
-  - success: docs name fail-forward behavior
-  - likely_fail: README still says all rejects keep goal open
-  - sneaky_fail: docs imply subagent evidence was accepted when it timed out
-  - UAT: [README.md](/home/wassname/.pi/agent/git/github.com/wassname/pi-plan/README.md)
-- [x] T4 (R5): Copy oracle's reliability shape for model/progress.
-  - verify: `npm run typecheck`
-  - success: `CompleteGoal` passes the current session model to the judge when no override is set, never spawns without `--model`, and streamed judge deltas are surfaced through `onUpdate`
-  - likely_fail: judge still runs without `--model`
-  - sneaky_fail: user sees no progress for several minutes and kills a working judge
-  - UAT: [src/index.ts](/home/wassname/.pi/agent/git/github.com/wassname/pi-plan/src/index.ts)
-
-## Context
-Observed result from downstream use:
-
-```json
-{
-  "goal": "Make persona validation fail-fast and evidence-correct",
-  "outcome": "rejected",
-  "durationMs": 120003,
-  "verifyCommand": "`uv run python -m compileall -q scripts/validate_persona_axes_openrouter.py`",
-  "reasoning": "VERDICT: reject\nmissing: judge timed out after 120s",
-  "isError": true
-}
-```
-
-Interpretation: latest surfaced output proves the internal judge timed out. It does not prove the verify command passed, though earlier logs indicated that pattern.
-
-## Log
-- 2026-06-29  current `runJudge` uses raw `pi -p --no-session` output plus ANSI stripping; oracle uses `--mode json` and parses message events, which is likely more reliable.
-- 2026-06-29  unset `/goals judge` spawns the judge without `--model`, so Pi resolves its configured default model; do not describe this as the current session model.
-- 2026-06-29  timeout/transport failure now maps to `accepted_inconclusive`, preserving partial output in reasoning when available.
-- 2026-06-29  fresh-eyes review found loose `/accept/i` verdict parsing and caller-abort fail-forward risk; fixed exact verdict parsing and made caller abort reject.
-- 2026-06-29  oracle comparison suggests the important reliability pieces are explicit model selection, JSON streaming, live partial output, and no short wrapper timeout; updated CompleteGoal to use the current session model when visible, never spawn without `--model`, stream throttled progress, and wait 600s.
-
-## TODO
-- Consider making `CompleteGoal` expose `verifyExitCode: 0` and `judgeOutcome` separately in details.
-
-## Errors
-| Task | Error | Resolution |
-|------|-------|------------|

package.json

@@ -1,7 +1,7 @@
 {
   "name": "@wassname2/pi-goals",
-  "version": "0.1.1",
-  "description": "One .pi/plan.md: set goals in plan mode, work them, sign off only when a read-only judge checks the evidence.",
+  "version": "0.3.3",
+  "description": "Set goals in plan.md; a smart supervisor guides cheap worker subagents through long autonomous sessions until your goals are signed off, with every agent's pane visible to you.",
   "author": "wassname",
   "license": "MIT",
   "type": "module",
@@ -18,15 +18,17 @@
     "proof",
     "uat",
     "evidence",
-    "judge"
+    "supervisor",
+    "herdr"
   ],
-  "dependencies": {
-    "@earendil-works/pi-coding-agent": "^0.79.0",
+  "peerDependencies": {
+    "@earendil-works/pi-coding-agent": ">=0.85.1 <1.0.0",
     "@earendil-works/pi-tui": "*",
-    "@sinclair/typebox": "latest"
+    "typebox": "*"
   },
   "files": [
     "src",
+    "agents",
     "README.md"
   ],
   "publishConfig": {
@@ -35,22 +37,42 @@
   "scripts": {
     "build": "tsc",
     "prepublishOnly": "npm run lint && npm run typecheck && npm run test",
-    "test": "vitest run",
-    "test:watch": "vitest",
+    "test": "vitest run --dir test",
+    "test:rpc": "vitest run --dir test rpc-review.test.ts",
+    "test:watch": "vitest --dir test",
     "typecheck": "tsc --noEmit",
     "lint": "biome check src/ test/",
     "lint:fix": "biome check --fix src/ test/"
   },
   "devDependencies": {
-    "@types/node": "^20.0.0",
-    "typescript": "^5.0.0",
     "@biomejs/biome": "^2.4.8",
+    "@earendil-works/pi-coding-agent": "0.85.1",
+    "@earendil-works/pi-tui": "^0.85.1",
+    "@types/node": "^20.0.0",
+    "typebox": "^1.3.7",
+    "typescript": "^5.0.0",
     "vitest": "^4.0.18"
   },
   "pi": {
     "extensions": [
-      "./src/index.ts"
+      "./src/index.ts",
+      "./node_modules/pi-subagents/src/index.ts",
+      "./node_modules/pi-intercom/index.ts",
+      "./node_modules/pi-schedule-prompt/src/index.ts"
     ],
-    "image": "https://cdn.jsdelivr.net/gh/wassname/pi-goals@main/media/screenshot.png"
-  }
+    "image": "https://github.com/user-attachments/assets/35feaa15-f022-4491-bcc2-fc31cb878a9f",
+    "skills": [
+      "./node_modules/pi-intercom/skills"
+    ]
+  },
+  "dependencies": {
+    "pi-subagents": "git+https://github.com/edxeth/pi-subagents.git#953c6f6d2fc7d8a5c956c30cd77c51bad697c2a4",
+    "pi-intercom": "0.13.0",
+    "pi-schedule-prompt": "0.4.1"
+  },
+  "bundledDependencies": [
+    "pi-subagents",
+    "pi-intercom",
+    "pi-schedule-prompt"
+  ]
 }

scripts/check-judge-footprint.sh

@@ -1,43 +0,0 @@
-#!/usr/bin/env bash
-# Replaces the stale FIXME(side-effect) claim in src/index.ts with a checked fact.
-#
-# The claim was: "pi -p --no-session clones the repo into the PARENT of cwd, leaving a stale
-# directory." Reproducing the exact sign-off judge invocation (pi --mode json -p --no-session,
-# read-only tools, edit/write excluded, cwd = here) shows it does not. This script makes that
-# reproducible: it runs the invocation, requires pi to actually reach agent_end (so a pass is not
-# vacuous), and asserts the parent-of-cwd listing is byte-identical before and after.
-#
-# Exit 0 = judge leaves no clone in the parent. Exit 1 = either pi did not run, or it polluted.
-# Run by hand; re-run as the rigorous sign-off check (the judge has bash and runs this itself).
-set -u
-
-PARENT="$(cd "$PWD/.." && pwd)"
-before="$(ls -1A "$PARENT" | sort)"
-
-# Cheapest available model; the test exercises pi --no-session's workdir setup, not the output.
-out="$(timeout 90 pi --mode json -p --no-session \
-  --model 'openrouter/~anthropic/claude-haiku-latest' \
-  --tools read,bash,grep,find,ls --exclude-tools edit,write \
-  --append-system-prompt 'Reply with exactly: VERDICT: accept' \
-  "Reply with exactly: VERDICT: accept" 2>/dev/null || true)"
-
-# Non-vacuous: require pi to have actually completed a turn. A pass without this could mean pi
-# crashed instantly and never had the chance to clone -- which would prove nothing.
-if ! printf '%s' "$out" | grep -q '"type":"agent_end"'; then
-  echo "FAIL: pi --no-session did not reach agent_end; cannot confirm no-clone."
-  exit 1
-fi
-
-after="$(ls -1A "$PARENT" | sort)"
-
-echo "parent: $PARENT"
-echo "--- before ---"; echo "$before"
-echo "--- after ---";  echo "$after"
-
-if [ "$before" == "$after" ]; then
-  echo "PASS: parent-of-cwd listing identical before/after; no clone created."
-  exit 0
-fi
-echo "FAIL: parent-of-cwd listing changed. Diff (< before, > after):"
-diff <(printf '%s\n' "$before") <(printf '%s\n' "$after") | head -20
-exit 1

scripts/check-stale-fixmes.sh

@@ -1,14 +0,0 @@
-#!/usr/bin/env bash
-# Structural gate for the goal's `verify:` field. Cheap and deterministic: no API calls.
-# Confirms (a) neither stale FIXME tag remains in src/, (b) the footprint script exists, and
-# (c) the plan-injection heading prefix is still emitted. The rigorous runtime check (running
-# the footprint script) is the sign-off judge's job -- it has bash and re-runs the script itself.
-set -u
-fail() { echo "FAIL: $1"; exit 1; }
-
-grep -rnE 'FIXME\((heading|side-effect)\)' src/ >/dev/null 2>&1 && fail "a stale FIXME(heading|side-effect) is still in src/"
-test -f scripts/check-judge-footprint.sh || fail "scripts/check-judge-footprint.sh is missing"
-grep -q '\.pi/goals\.md:' src/prompts.ts || fail "the .pi/goals.md: heading prefix was dropped from src/prompts.ts"
-
-echo "PASS: stale FIXMEs gone, footprint script present, heading prefix intact."
-exit 0

scripts/inconclusive-fail-forward.diff

@@ -1,104 +0,0 @@
-diff --git a/README.md b/README.md
-index ce4056a..5485002 100644
---- a/README.md
-+++ b/README.md
-@@ -145,9 +145,9 @@ else is the agent editing the file. It reads the goal's `evidence:` block from `
-    reasoning comes back in the result.
- 
- The judge defaults to the current session model and streams partial output while it runs. If the
--current model is not visible to the extension, `CompleteGoal` does not fall back to Pi's implicit
--default; it signs off as `judge inconclusive` and tells you to set `/goals judge <provider/model>`.
--Point it at another model for an independent cross-family check.
-+session model is not visible to the extension, the `--model` flag is omitted and pi uses its own
-+configured default, so the judge always runs. `/goals judge <provider/model>` is an optional override
-+for an independent cross-family check; never required.
- 
- ## Prompts
- 
-diff --git a/src/index.ts b/src/index.ts
-index 9eb2a16..14784f7 100644
---- a/src/index.ts
-+++ b/src/index.ts
-@@ -326,7 +326,7 @@ export default function piGoalsExtension(pi: ExtensionAPI): void {
- 				durationMs,
- 				verifyCommand: goal.verify ?? undefined,
- 				verifyExitCode: outcome.kind === "verify_failed" ? outcome.exitCode : undefined,
--				judgeModel: judgeModel ?? "no explicit judge model",
-+				judgeModel: judgeModel ?? "pi default",
- 				reasoning,
- 				isError: res.isError,
- 			};
-@@ -522,14 +522,6 @@ async function decideSignOff(
- 			};
- 		}
- 	}
--	if (!judgeModel) {
--		const reason = "no explicit judge model available; set /goals judge <provider/model>";
--		return {
--			outcome: { kind: "accepted_inconclusive", reason },
--			reasoning: `VERDICT: inconclusive\nreason: ${reason}`,
--			durationMs: Date.now() - startedAt,
--		};
--	}
- 	const verdict = await runJudge(goal, evidence, paths, verifyResult, judgeModel, cwd, signal, onUpdate);
- 	const outcome: SignOff =
- 		verdict.kind === "accepted"
-@@ -573,13 +565,25 @@ type JudgeResult =
- 	| { kind: "rejected"; missing: string; reasoning: string; durationMs: number }
- 	| { kind: "inconclusive"; reason: string; reasoning: string; durationMs: number };
- 
-+/** Stage 2: a read-only pi subprocess inspects the evidence against the repo and returns a verdict. */
-+/** Build the pi argv for the read-only judge. `--model` is omitted when no explicit/session model is
-+ *  set, so pi falls back to its configured default -- the judge always runs, never pre-emptively
-+ *  fails as "no model". Exported for a unit test that locks this invariant (an empty `--model ""`
-+ *  would make every sign-off silently inconclusive). */
-+export function buildJudgeArgs(judgeModel: string | null): string[] {
-+	const args = ["--mode", "json", "-p", "--no-session"];
-+	if (judgeModel) args.push("--model", judgeModel);
-+	args.push("--tools", JUDGE_TOOLS.join(","), "--exclude-tools", JUDGE_BLOCKED_TOOLS.join(","), "--append-system-prompt", evidenceJudgeSystem);
-+	return args;
-+}
-+
- /** Stage 2: a read-only pi subprocess inspects the evidence against the repo and returns a verdict. */
- async function runJudge(
- 	goal: Goal,
- 	evidence: string,
- 	paths: string[],
- 	verifyResult: { command: string; exitCode: number; outputTail: string } | null,
--	judgeModel: string,
-+	judgeModel: string | null,
- 	cwd: string,
- 	signal: AbortSignal | undefined,
- 	onUpdate?: (partial: { content: Array<{ type: "text"; text: string }>; details: SignOffDetails }) => void,
-@@ -600,14 +604,14 @@ async function runJudge(
- 		evidence,
- 		paths,
- 	});
--	const args = ["--mode", "json", "-p", "--no-session", "--model", judgeModel, "--tools", JUDGE_TOOLS.join(","), "--exclude-tools", JUDGE_BLOCKED_TOOLS.join(","), "--append-system-prompt", evidenceJudgeSystem];
-+	const args = buildJudgeArgs(judgeModel);
- 	args.push(task);
- 
- 	emit("spawning", `Spawning read-only judge for: ${goal.subject}`);
- 	const inv = getPiInvocation(args);
--	// FIXME(side-effect): pi -p --no-session clones the repo into the PARENT of cwd (so alongside
--	// the working dir), leaving a stale directory. The judge should run in a temp dir or inside the
--	// existing repo checkout so it doesn't pollute the user's workspace.
-+	// The judge runs in-place against this checkout (cwd is passed to spawn and the read-only tools
-+	// read from it); pi --no-session does not clone into the parent. Proven and re-checked by
-+	// scripts/check-judge-footprint.sh, which reproduces this invocation and asserts no parent clone.
- 	const judge = await new Promise<{ output: string; error?: string; aborted?: boolean }>((resolve) => {
- 		let settled = false;
- 		let stdoutBuffer = "";
-diff --git a/src/prompts.ts b/src/prompts.ts
-index 03faea8..3b8d270 100644
---- a/src/prompts.ts
-+++ b/src/prompts.ts
-@@ -117,8 +117,6 @@ export function planInjection(p: {
-   counts: { done: number; open: number };
- }): string {
-   if (!p.activeGoal) {
--    // FIXME(heading): user wants the heading to show ".pi/goals.md: <title>" so the filename is explicit
--    // even in the injection. Currently says "Goals (goals.md):" which is close but not the same.
-     return `.pi/goals.md: ${p.title}\nNo active goal. ${p.counts.open} open, ${p.counts.done} done. Pick the next goal (set its checkbox to [/]) or run /goals.`;
-   }
-   const subtasks = p.activeGoal.openSubtasks.length

scripts/prepare-trial.mjs

@@ -0,0 +1,35 @@
+// Pi/OpenAI. Prepare an isolated trial; never launches/reloads an existing session.
+import { copyFileSync, existsSync, mkdirSync, mkdtempSync, readFileSync, writeFileSync } from 'node:fs';
+import { dirname, join, resolve } from 'node:path';
+import { fileURLToPath, pathToFileURL } from 'node:url';
+import { execFileSync } from 'node:child_process';
+import { homedir, tmpdir } from 'node:os';
+const repo = resolve(dirname(fileURLToPath(import.meta.url)), '..');
+const sdkRoot = process.argv[2];
+const noSandbox = process.argv.includes('--no-sandbox');
+if (!sdkRoot) throw new Error('Usage: node scripts/prepare-trial.mjs INSTALLED_PI_ROOT');
+const revision = execFileSync('git', ['-C', repo, 'rev-parse', 'HEAD'], {encoding:'utf8'}).trim();
+const root = mkdtempSync(join(tmpdir(), 'goals-edxeth-trial-'));
+const cwd = join(root, 'project'); const agentDir = join(root, 'agent');
+mkdirSync(cwd); mkdirSync(agentDir, {mode:0o700}); mkdirSync(join(agentDir,'agents'));
+const sourceAgent = process.env.PI_CODING_AGENT_DIR || join(homedir(),'.pi','agent');
+const sourceSettings = JSON.parse(readFileSync(join(sourceAgent,'settings.json'),'utf8'));
+const sdk = await import(pathToFileURL(join(sdkRoot,'dist/index.js')).href);
+const settings = sdk.SettingsManager.create(repo, sourceAgent, { projectTrusted:false });
+const manager = new sdk.DefaultPackageManager({cwd:repo, agentDir:sourceAgent, settingsManager:settings});
+const packages = manager.listConfiguredPackages().filter((p) => p.scope !== 'project' && !/^\/\//.test(p.source));
+const retained = packages.filter((p) => !/pi-subagents|pi-goals|pi-intercom|pi-schedule-prompt/.test(p.source));
+for (const p of retained) if (!p.installedPath) throw new Error(`Missing installed package: ${p.source}`);
+writeFileSync(join(agentDir,'settings.json'), JSON.stringify({...sourceSettings, packages:[...retained.map((p)=>p.installedPath), repo]},null,2));
+// Private copies, not symlinks: a trial OAuth refresh must not write the active auth file.
+for (const file of ['auth.json','models.json']) if (existsSync(join(sourceAgent,file))) copyFileSync(join(sourceAgent,file),join(agentDir,file));
+const workerDefinition = readFileSync(join(repo,'agents/goals-worker.md'),'utf8');
+writeFileSync(join(agentDir,'agents/goals-worker.md'), noSandbox ? workerDefinition.replace('mode: interactive', 'mode: interactive\nflags: --no-sandbox') : workerDefinition);
+execFileSync('git',['init','--quiet',cwd]);
+writeFileSync(join(cwd,'AGENTS.md'), 'Isolated functional trial. Work only in this project. Do not operate other Herdr panes, use live research sessions, or change global settings. Preserve evidence. The main chat supervises; the goals-worker implements.\n');
+writeFileSync(join(cwd,'.gitignore'), 'evidence/\n');
+const manifest={root,cwd,agentDir,repo,revision,noSandbox,retainedPackages:retained.map((p)=>p.source),replacedPackages:packages.filter((p)=>!retained.includes(p)).map((p)=>p.source)};
+writeFileSync(join(root,'manifest.json'),JSON.stringify(manifest,null,2));
+const quote=(s)=>`'${s.replaceAll("'", "'\\''")}'`;
+writeFileSync(join(root,'start.zsh'), `#!/usr/bin/env zsh\nset -e\ncd ${quote(cwd)}\nexport PI_CODING_AGENT_DIR=${quote(agentDir)}\nexport PI_SUBAGENT_MUX=herdr\nexport PI_ORCHESTRATOR_MODE=0\nexec pi --approve${noSandbox ? ' --no-sandbox' : ''}\n`,{mode:0o700});
+console.log(JSON.stringify({root,cwd,agentDir,start:join(root,'start.zsh'),manifest:join(root,'manifest.json')},null,2));

scripts/session-usage.mjs

@@ -0,0 +1,61 @@
+// Pi/OpenAI: Sum recorded requests, not context occupancy; do not read message text.
+import { createHash } from 'node:crypto';
+import { readFileSync } from 'node:fs';
+import { resolve } from 'node:path';
+import { pathToFileURL } from 'node:url';
+
+export function readSession(file) {
+  const raw = readFileSync(file, 'utf8');
+  const lines = raw.split('\n');
+  const tail = lines.pop();
+  let trailingPartial = false;
+  if (tail) {
+    try { JSON.parse(tail); lines.push(tail); }
+    catch { trailingPartial = true; }
+  }
+  return { file: resolve(file), sha256: createHash('sha256').update(raw).digest('hex'), trailingPartial,
+    entries: lines.filter(Boolean).map(JSON.parse) };
+}
+
+export function summarize(entries, since, until) {
+  const start = Date.parse(since), end = Date.parse(until);
+  if (!Number.isFinite(start) || !Number.isFinite(end) || start > end) throw new Error('Invalid time interval');
+  const rows = entries.filter(e => e.type === 'message' && e.message.role === 'assistant' && Date.parse(e.timestamp) >= start && Date.parse(e.timestamp) <= end);
+  const totals = { calls: 0, input: 0, cacheRead: 0, cacheWrite: 0, output: 0, totalTokens: 0 };
+  const models = new Map();
+  let missingUsage = 0;
+  for (const e of rows) {
+    const m = e.message;
+    if (!m.usage) { missingUsage++; continue; }
+    const model = `${m.provider}/${m.model}`;
+    if (!models.has(model)) models.set(model, { model, ...totals, calls: 0, input: 0, cacheRead: 0, cacheWrite: 0, output: 0, totalTokens: 0 });
+    const group = models.get(model);
+    totals.calls++; group.calls++;
+    for (const key of ['input', 'cacheRead', 'cacheWrite', 'output', 'totalTokens']) {
+      const value = m.usage[key];
+      if (!Number.isFinite(value) || value < 0) throw new Error(`Invalid usage.${key} in entry ${e.id}`);
+      totals[key] += value; group[key] += value;
+    }
+  }
+  return { ...totals, missingUsage, firstRequest: rows[0]?.timestamp ?? null,
+    lastRequest: rows.at(-1)?.timestamp ?? null, models: [...models.values()] };
+}
+
+export function report(supervisor, worker, until = new Date().toISOString()) {
+  const boundary = supervisor.entries.findLast(e => e.type === 'custom' && e.customType === 'pi-goals-main-supervisor-v1' && e.data.mode === 'planning' && !e.data.child);
+  if (!boundary) throw new Error('No recorded planning start in supervisor session');
+  const since = boundary.timestamp;
+  const sessions = [supervisor, worker].map((session, i) => ({
+    role: i === 0 ? 'supervisor' : 'worker', file: session.file, sha256: session.sha256,
+    trailingPartial: session.trailingPartial, ...summarize(session.entries, since, until),
+  }));
+  return { since, until, elapsedHours: (Date.parse(until) - Date.parse(since)) / 3600000,
+    boundaryEntry: boundary.id, plan: boundary.data.plan, sessions,
+    scope: 'Recorded assistant usage since latest planning entry, including abandoned branches and repeated cached context. Excludes earlier inherited history, in-flight requests, subprocess API usage and unrecorded compaction calls. Output includes reasoning where the provider includes it; reasoning is not added twice.' };
+}
+
+if (process.argv[1] && import.meta.url === pathToFileURL(resolve(process.argv[1])).href) {
+  const [supervisor, worker] = process.argv.slice(2);
+  if (!supervisor || !worker || process.argv.length !== 4) throw new Error('Usage: node scripts/session-usage.mjs SUPERVISOR.jsonl WORKER.jsonl');
+  console.log(JSON.stringify(report(readSession(supervisor), readSession(worker)), null, 2));
+}

scripts/stale-fixme-removal.diff

@@ -1,30 +0,0 @@
-diff --git a/src/index.ts b/src/index.ts
-index 9eb2a16..cdd7b45 100644
---- a/src/index.ts
-+++ b/src/index.ts
-@@ -605,9 +605,9 @@ async function runJudge(
- 
- 	emit("spawning", `Spawning read-only judge for: ${goal.subject}`);
- 	const inv = getPiInvocation(args);
--	// FIXME(side-effect): pi -p --no-session clones the repo into the PARENT of cwd (so alongside
--	// the working dir), leaving a stale directory. The judge should run in a temp dir or inside the
--	// existing repo checkout so it doesn't pollute the user's workspace.
-+	// The judge runs in-place against this checkout (cwd is passed to spawn and the read-only tools
-+	// read from it); pi --no-session does not clone into the parent. Proven and re-checked by
-+	// scripts/check-judge-footprint.sh, which reproduces this invocation and asserts no parent clone.
- 	const judge = await new Promise<{ output: string; error?: string; aborted?: boolean }>((resolve) => {
- 		let settled = false;
- 		let stdoutBuffer = "";
-diff --git a/src/prompts.ts b/src/prompts.ts
-index 03faea8..3b8d270 100644
---- a/src/prompts.ts
-+++ b/src/prompts.ts
-@@ -117,8 +117,6 @@ export function planInjection(p: {
-   counts: { done: number; open: number };
- }): string {
-   if (!p.activeGoal) {
--    // FIXME(heading): user wants the heading to show ".pi/goals.md: <title>" so the filename is explicit
--    // even in the injection. Currently says "Goals (goals.md):" which is close but not the same.
-     return `.pi/goals.md: ${p.title}\nNo active goal. ${p.counts.open} open, ${p.counts.done} done. Pick the next goal (set its checkbox to [/]) or run /goals.`;
-   }
-   const subtasks = p.activeGoal.openSubtasks.length

src/index.ts

@@ -1,527 +1,446 @@
-/**
- * pi-goals v2 — plan mode drafts goals into one .pi/plan.md, the agent works them with its normal
- * Edit tool, and a fresh read-only judge signs each goal off through the one blessed tool,
- * CompleteGoal.
- *
- * The v1 lesson: the parser existed so TypeScript could read plan.md, but almost every reader is a
- * model. So v2 has NO parser and no schema. The harness does exactly three things for a
- * cooperative-but-confused model:
- *   1. memory  — a transient re-send of the plan, never persisted, on two triggers: the plan went
- *                stale for STALE_TURNS turns (send the working set above ## Log), or the session
- *                started / compacted (send the whole file, appendix included). v2 sent the whole
- *                file every turn; pi-tasks tried that and deleted it as "wallpaper noise that
- *                trains the model to ignore the task block" (tintinweb/pi-tasks CHANGELOG.md:149),
- *                and the always-present CompleteGoal description carries the contract instead.
- *   2. format  — a skeleton convention taught in planDrafting (prompts.ts), not validated
- *   3. eyes    — CompleteGoal spawns a strictly read-only pi subprocess (--no-session, no bash)
- *                that gets the whole plan file plus the claimed goal, finds the goal itself
- *                (tolerates wording drift), checks the evidence (including the agent's saved
- *                verify output) against the repo, and returns VERDICT: accept|reject
- *
- * The judge subsumes what v1 did in code: goal matching (no findGoal), evidence validation (a
- * placeholder gets rejected in words), and format reading. The extension's only
- * writes are the sign-off: append a log line to ## Log (the audit trail) and tick the goal [x] when
- * an exact goal line matches (on drift the agent ticks, and the result says so). A hand-tick
- * without a matching tool-written log line is visible in the diff either way.
- *
- * Judge ran but failed/errored/timed out, or returned no VERDICT line => accepted_inconclusive: the
- * working agent is never blocked on judge infra; the log line says the judge ran but failed. There
- * is no pre-emptive "no model" path -- a null judgeModel just omits --model so pi's configured
- * default runs the judge, so inconclusive always means "ran but failed", never "couldn't start".
- *
- * All model-facing text lives in prompts.ts, in flow order.
- */
-
-import { spawn, spawnSync } from "node:child_process";
-import { existsSync, mkdirSync, readFileSync, renameSync, writeFileSync } from "node:fs";
-import { basename, join, resolve } from "node:path";
+// Pi/OpenAI: Plan and supervise in the main chat; delegate implementation to a visible worker.
+import { createHash } from "node:crypto";
+import { type FSWatcher, mkdirSync, readFileSync, watch, writeFileSync } from "node:fs";
+import { dirname, isAbsolute, join, resolve } from "node:path";
 import type { ExtensionAPI, ExtensionContext } from "@earendil-works/pi-coding-agent";
-import { Type } from "@sinclair/typebox";
-import { completeGoalDescription, completeGoalParamDescription, judgeSystem, judgeUser, planDrafting, reminder, resync } from "./prompts.js";
+import { Type } from "typebox";
+import { foldPlan, GOAL_LINE } from "./plan.js";
+import { planViews } from "./plan-view.js";
+import {
+	attachGoalPlanDescription,
+	attachNotice,
+	childPlanAttached,
+	childPlanRole,
+	completeGoalDescription,
+	completionLog,
+	completionResult,
+	discuss,
+	emptyEvidence,
+	evidenceUnavailable,
+	goalToolBlocked,
+	manualReview,
+	messages,
+	pausedRole,
+	pauseExitNotice,
+	planChangedReview,
+	planContext,
+	planDocument,
+	planning,
+	planningSeed,
+	planUnavailable,
+	readyApproved,
+	removeGoalSchedule,
+	resumeNotice,
+	scheduleCheckIn,
+	soloNotice,
+	soloRole,
+	supervisor,
+	upkeep,
+} from "./prompts.js";
 
-const STATE = "pi-goals-state";
-const STATUS_KEY = "pi-goals";
-const WIDGET_KEY = "pi-goals-widget";
-const PLAN_REL = ".pi/plan.md";
-// Judge toolset: strictly read-only, NO bash -- the judge can never execute or mutate anything, and
-// in particular never re-runs a verify command (which may be a 10-hour training job). The agent runs
-// verify itself and saves the output as evidence; the judge reads it. Names match pi's tool registry.
-const JUDGE_TOOLS = ["read", "grep", "find", "ls"];
-const JUDGE_BLOCKED_TOOLS = ["edit", "write"];
-const JUDGE_TIMEOUT_MS = 600_000;
-// Plan mode is read-only by convention AND a light gate: edit/write are blocked (except plan.md,
-// the deliverable). bash stays open — the prompt says don't mutate; guide, don't gate (spec D3).
-const PLAN_MODE_BLOCKED_TOOLS = ["edit", "write"];
-// Turns the plan may go untouched before it is re-sent. pi-tasks uses 4, or 2 while something is in
-// progress; here every goal is "in progress", so 2.
-const STALE_TURNS = 2;
-
-// A checkbox line beginning "goal:", for the widget and the "any goals open?" reminder condition.
-// Everything else reads the file as prose.
-const GOAL_LINE = /^\s*(?:\d+\.|[-*])\s*\[([ xX/-])\]\s*goal:\s*(.*)$/i;
-// An indented checkbox line that isn't a goal: a subtask. Only the widget reads these, so the human
-// sees the next action and not just the goal -- this file IS the task list.
-const SUBTASK_LINE = /^\s+(?:\d+\.|[-*])\s*\[([ xX/-])\]\s*(.*)$/;
-// The fold. Above it: the working set that gets re-sent. Below it: durable memory.
-const FOLD_LINE = /^##\s+Log\s*$/im;
+const STATE = "pi-goals-main-supervisor-v1";
+const WORKER = "goals-worker";
+type Mode = "chat" | "planning" | "supervising" | "paused" | "solo";
 type GoalStatus = "open" | "active" | "done" | "cancelled";
-const CHAR_TO_STATUS: Record<string, GoalStatus> = { " ": "open", "/": "active", x: "done", "-": "cancelled" };
-
-function scanGoals(plan: string): Array<{ status: GoalStatus; subject: string; line: number }> {
-	const goals: Array<{ status: GoalStatus; subject: string; line: number }> = [];
-	plan.split("\n").forEach((line, i) => {
-		const m = GOAL_LINE.exec(line);
-		if (m) goals.push({ status: CHAR_TO_STATUS[m[1].toLowerCase()] ?? "open", subject: m[2].trim(), line: i });
+interface State {
+	mode: Mode;
+	plan?: string;
+	worker?: { id?: string; sessionFile: string };
+	helpers: { id?: string; sessionFile: string }[];
+	workerStopped?: boolean;
+	signoffs: Record<string, { evidence: string[]; observation: string }>;
+	child?: boolean;
+}
+const initial = (): State => ({ mode: "chat", helpers: [], signoffs: {} });
+const digest = (text: string) => createHash("sha256").update(text).digest("hex");
+const key = (text: string) => text.trim().toLowerCase();
+function goals(text: string) {
+	return foldPlan(text).split("\n").flatMap((line, index) => {
+		const match = GOAL_LINE.exec(line);
+		if (!match) return [];
+		const box = match[1].toLowerCase();
+		return [{ subject: match[2].trim(), status: (box === "x" ? "done" : box === "/" ? "active" : box === "-" ? "cancelled" : "open") as GoalStatus, index }];
 	});
-	return goals;
 }
+const result = (text: string) => ({ content: [{ type: "text" as const, text }], details: {} });
 
-/** The working set: everything above "## Log". Log, Learnings and Appendix below it are durable
- *  memory -- unlimited, read on demand, pushed back only by a resync. Exported for the unit test. */
-export function foldPlan(plan: string): string {
-	const m = FOLD_LINE.exec(plan);
-	return (m ? plan.slice(0, m.index) : plan).trimEnd();
-}
-
-/** Open subtasks under the goal on line `goalLine`, up to the next goal line. */
-export function openSubtasks(plan: string, goalLine: number): string[] {
-	const lines = plan.split("\n");
-	const out: string[] = [];
-	for (let i = goalLine + 1; i < lines.length; i++) {
-		if (GOAL_LINE.test(lines[i])) break;
-		const m = SUBTASK_LINE.exec(lines[i]);
-		if (m && (m[1] === " " || m[1] === "/")) out.push(m[2].trim());
-	}
-	return out;
-}
-
-interface PlanState {
-	isPlanMode: boolean;
-	/** Optional model ref for the sign-off judge; unset => current session model, else pi's default. */
-	judgeModel: string | null;
-}
-
-export default function piGoalsExtension(pi: ExtensionAPI): void {
-	let state: PlanState = { isPlanMode: false, judgeModel: null };
-	// Reminder cadence (pi-tasks style): the plan is re-sent only after it has gone untouched for
-	// STALE_TURNS turns, and editing it resets the clock -- an agent that is maintaining the file
-	// doesn't need to be told to. In-memory, like pi-tasks: a new session starts fresh.
-	let turnsStale = 0;
-	let lastSeenPlan = "";
-	// Set on session start and after a compaction; drained by the next LLM call, which then carries
-	// the WHOLE file (appendix included) instead of just the working set.
-	let resyncReason: string | null = "New session.";
-
-	const planPath = (ctx: ExtensionContext) => join(ctx.cwd, ".pi", "plan.md");
-	const readPlan = (ctx: ExtensionContext): string => (existsSync(planPath(ctx)) ? readFileSync(planPath(ctx), "utf-8") : "");
-	const writePlan = (ctx: ExtensionContext, content: string): void => {
-		mkdirSync(join(ctx.cwd, ".pi"), { recursive: true });
-		writeFileSync(planPath(ctx), content);
+export default function mainSupervisor(pi: ExtensionAPI) {
+	let state = initial();
+	let generation = 0;
+	let workerRevision = 0;
+	let pendingLaunches = 0;
+	let notice = true;
+	let planWatcher: FSWatcher | undefined;
+	let planEditTimer: ReturnType<typeof setTimeout> | undefined;
+	let planHash = "";
+	const childEnvironment = process.env.PI_SUBAGENT_AGENT === WORKER;
+	const save = () => pi.appendEntry(STATE, state);
+	// Missing, empty and failed reads are unavailable snapshots, never an empty authoritative plan.
+	const readPlan = () => {
+		try {
+			if (!state.plan) throw new Error(messages.noPlan);
+			const text = readFileSync(state.plan, "utf8");
+			if (!text.trim()) throw new Error(messages.emptyPlan);
+			return { text };
+		} catch (error) { return { error: planUnavailable(state.plan, error) }; }
 	};
-
-	function persist(): void {
-		pi.appendEntry<PlanState>(STATE, state);
-	}
-
-	function updateWidget(ctx: ExtensionContext): void {
-		if (state.isPlanMode) {
-			ctx.ui.setStatus(STATUS_KEY, ctx.ui.theme.fg("warning", "planning"));
-			ctx.ui.setWidget(WIDGET_KEY, [`pi-goals: drafting goals in ${PLAN_REL}`]);
+	const planText = () => {
+		const snapshot = readPlan();
+		if (snapshot.text === undefined) throw new Error(snapshot.error);
+		return snapshot.text;
+	};
+	let turnsStale = 0;
+	let lastWorkingSet = "";
+	const checkIn = (ctx: ExtensionContext) => scheduleCheckIn(ctx.sessionManager.getSessionId(), state.plan ?? "");
+	const hasScheduleTool = () => pi.getAllTools().some((tool) => tool.name === "schedule_prompt");
+	const notedPlanValue = (prefix: string) => {
+		const snapshot = readPlan();
+		if (snapshot.text === undefined) return null;
+		const m = new RegExp(`^\\-\\s*${prefix}:\\s*(.+)$`, "im").exec(foldPlan(snapshot.text));
+		return m?.[1]?.trim() ?? null;
+	};
+	function refresh(ctx: ExtensionContext) {
+		if (state.mode === "chat") { ctx.ui.setStatus("goals", undefined); ctx.ui.setWidget("goals", undefined); return; }
+		const snapshot = readPlan();
+		if (snapshot.text === undefined) {
+			ctx.ui.setStatus("goals", snapshot.error);
+			ctx.ui.setWidget("goals", [snapshot.error]);
 			return;
 		}
-		const goals = scanGoals(readPlan(ctx));
-		if (goals.length === 0) {
-			ctx.ui.setStatus(STATUS_KEY, undefined);
-			ctx.ui.setWidget(WIDGET_KEY, undefined);
-			return;
+		const items = goals(snapshot.text);
+		// Reopened/deleted/ambiguous goal identities lose their sign-off. Manual ticks remain claims.
+		for (const subject of Object.keys(state.signoffs)) {
+			const matches = items.filter((g) => key(g.subject) === subject);
+			if (matches.length !== 1 || matches[0].status !== "done") { delete state.signoffs[subject]; save(); }
 		}
-		const done = goals.filter((g) => g.status === "done").length;
-		ctx.ui.setStatus(STATUS_KEY, ctx.ui.theme.fg("accent", `◷ ${done}/${goals.length} goals`));
-		const mark: Record<GoalStatus, string> = { done: "✔", active: "▸", open: "◻", cancelled: "✗" };
-		// Only live goals get lines so finished work never pushes current work off screen. The active
-		// goal also shows its open subtasks: this file is the task list, so the widget is the task list.
-		const plan = readPlan(ctx);
-		const lines = [ctx.ui.theme.fg("muted", PLAN_REL)];
-		for (const g of goals.filter((g) => g.status === "active" || g.status === "open")) {
-			lines.push(`${mark[g.status]} ${g.subject}`);
-			if (g.status === "active") lines.push(...openSubtasks(plan, g.line).slice(0, 3).map((s) => ctx.ui.theme.fg("muted", `   ◦ ${s}`)));
+		const accepted = items.filter((g) => g.status === "done" && state.signoffs[key(g.subject)]).length;
+		ctx.ui.setStatus("goals", `goals: ${state.child ? "worker" : state.mode} | ${accepted}/${items.length} reviewed`);
+		const mark = (status: GoalStatus, signed: boolean) => status === "done" ? (signed ? "✓" : "?") : status === "active" ? "▸" : status === "cancelled" ? "✗" : "○";
+		const lines: string[] = items.map((g) => `${mark(g.status, Boolean(state.signoffs[key(g.subject)]))} ${g.subject}`);
+		if (items.some((g) => g.status === "done" && !state.signoffs[key(g.subject)])) lines.push("? = completion claim; parent review still required");
+		ctx.ui.setWidget("goals", lines);
+	}
+	function watchPlan(ctx: ExtensionContext) {
+		planWatcher?.close();
+		planWatcher = undefined;
+		clearTimeout(planEditTimer);
+		planEditTimer = undefined;
+		const snapshot = readPlan();
+		if (snapshot.text !== undefined) planHash = digest(planViews(snapshot.text).notify);
+		if (state.child || state.mode !== "supervising" || !state.plan) return;
+		const stamp = generation;
+		// Watch the directory so atomic plan replacement remains observable. This is an event hook:
+		// plan-change reviews, not another scheduled loop (the hourly job is schedule_prompt's). A
+		// short debounce coalesces bursts. Existing high-level plan views exclude maintenance
+		// (tasks/evidence/Log) while preserving requirement wording and goal checkbox claims.
+		try {
+			planWatcher = watch(dirname(state.plan), { persistent: false }, () => {
+			if (stamp !== generation) return;
+			if (planEditTimer) clearTimeout(planEditTimer);
+			planEditTimer = setTimeout(() => {
+				planEditTimer = undefined;
+				if (stamp !== generation || state.mode !== "supervising") return;
+				const snapshot = readPlan();
+				if (snapshot.text === undefined) { ctx.ui.notify(snapshot.error!, "warning"); return; }
+				refresh(ctx);
+				const hash = digest(planViews(snapshot.text).notify);
+				if (hash === planHash) return;
+				planHash = hash;
+				notice = true;
+				send(planChangedReview(state.plan!));
+			}, 150);
+		});
+		planWatcher.on("error", (error) => { planWatcher?.close(); planWatcher = undefined; ctx.ui.notify(`Plan monitoring failed: ${error.message}`, "error"); });
+		} catch (error) { ctx.ui.notify(`Plan monitoring unavailable: ${String(error)}`, "error"); }
+	}
+	function restore(ctx: ExtensionContext) {
+		generation++;
+		state = initial();
+		for (const entry of ctx.sessionManager.getBranch()) {
+			if (entry.type === "custom" && entry.customType === STATE) state = structuredClone(entry.data as State);
 		}
-		ctx.ui.setWidget(WIDGET_KEY, lines);
+		if (childEnvironment) {
+			state.child = true;
+			state.mode = "solo";
+			// Lineage-only workers attach the explicit task path using AttachGoalPlan.
+			save();
+		}
+		state.helpers ??= []; // sessions persisted before helper bookkeeping
+		notice = true;
+		turnsStale = 0;
+		lastWorkingSet = "";
+		refresh(ctx);
+		watchPlan(ctx);
+	}
+	function compatible() {
+		const tools = pi.getAllTools();
+		const properties = (name: string) => (tools.find((t) => t.name === name)?.parameters as { properties?: Record<string, unknown> } | undefined)?.properties;
+		return properties("subagent")?.title && properties("subagent")?.agent && properties("subagent_resume")?.sessionFile && properties("subagent_kill")?.id;
+	}
+	function send(content: string, triggerTurn = true) {
+		// sendMessage(triggerTurn:true) bypasses before_agent_start in Pi 0.85.1.
+		// A normal saved prompt prepares the current role before starting the turn.
+		if (triggerTurn) pi.sendUserMessage(`[pi-goals]\n${content}`, { deliverAs: "followUp" });
+		else pi.sendMessage({ customType: "pi-goals-supervision", content, display: true }, { deliverAs: "followUp", triggerTurn: false });
+	}
+	async function confirmOwnership(ctx: ExtensionContext, target: string, text: string, solo = true): Promise<boolean> {
+		if (pendingLaunches > 0) { ctx.ui.notify("A worker launch/resume is still pending; inspect its result before takeover.", "warning"); return false; }
+		const stamp = generation;
+		const revision = workerRevision;
+		const confirmation = solo ? "Worker confirmed stopped" : "Previous supervisor confirmed stopped";
+		const choice = await ctx.ui.select(solo ? "Confirm all other writers for the current and target plans are stopped (inspect /subagents and their panes). A missing handle is not proof. Take over in this session?" : "Confirm no other supervisor owns this plan. Preserve any existing worker session and reconnect rather than starting another writer.", [confirmation, "Cancel"]);
+		if (stamp !== generation || revision !== workerRevision) return false;
+		if (choice !== confirmation) return false;
+		if (readFileSync(target, "utf8") !== text) { ctx.ui.notify("Plan changed during takeover; confirm again.", "warning"); return false; }
+		return true;
+	}
+	function enterSolo(ctx: ExtensionContext) {
+		state.mode = "solo"; state.workerStopped = true;
+		generation++; notice = true; save(); refresh(ctx); watchPlan(ctx);
+		send(`${removeGoalSchedule(ctx.sessionManager.getSessionId())}\n\n${soloNotice(state.plan!)}`);
+	}
+	const help = "/goals new [initial idea] | review | ready | status | stop | resume | solo | exit | attach <plan.md> [solo] | model <model>\n/subagents opens the worker controls. Stop/exit pause this plan locally; worker termination must be confirmed through subagent_kill or its pane. No forced compaction or model switch; the worker pane's own model is chosen with /model in that pane. Hourly check-ins are one session-bound schedule_prompt job; plan-change reviews are the plan-watcher event hook.";
+	async function ready(ctx: ExtensionContext, menu: boolean) {
+		if (state.mode !== "planning") { ctx.ui.notify("Ready applies to a draft; use status or resume.", "warning"); return; }
+		const text = planText();
+		const items = goals(text);
+		if (!items.length || new Set(items.map((g) => key(g.subject))).size !== items.length) {
+			ctx.ui.notify("Write a plan with distinct '- [ ] goal: ...' subjects before Ready.", "warning"); return;
+		}
+		const stamp = generation;
+		if (menu) {
+			const choice = await ctx.ui.select(`Review ${state.plan}`, ["Ready", "Discuss", "Edit", "Cancel"]);
+			if (stamp !== generation || digest(planText()) !== digest(text)) { ctx.ui.notify("Plan changed during review. Review it again.", "warning"); return; }
+			if (choice === "Discuss") { send(discuss); return; }
+			if (choice === "Edit") {
+				const edited = await ctx.ui.editor("Edit goal plan", text);
+				if (edited !== undefined && stamp === generation && planText() === text && state.plan) { writeFileSync(state.plan, edited); refresh(ctx); }
+				return;
+			}
+			if (choice !== "Ready") return;
+		}
+		if (!compatible()) { ctx.ui.notify("Requires edxeth/pi-subagents 2.9.x, not nicobailon/pi-subagents. Draft preserved; /goals solo is available.", "error"); return; }
+		state.mode = "supervising"; generation++; notice = true; save(); refresh(ctx); watchPlan(ctx);
+		send(`${checkIn(ctx)}\n\n${readyApproved(WORKER, state.plan!, state.worker?.sessionFile, text, ctx.sessionManager.getSessionId())}`);
 	}
 
-	// --- /goals: enter plan mode (or clear / set judge) --------------------------------------------
+	pi.on("session_start", (_e, ctx) => restore(ctx));
+	pi.on("session_tree", (_e, ctx) => restore(ctx));
+	pi.on("session_shutdown", () => { generation++; planWatcher?.close(); planWatcher = undefined; clearTimeout(planEditTimer); planEditTimer = undefined; });
+	pi.on("session_compact", () => { notice = true; });
+	pi.on("turn_end", (_event, ctx) => {
+		if (!["supervising", "solo"].includes(state.mode)) return;
+		const snapshot = readPlan();
+		if (snapshot.text === undefined) { notice = true; return; }
+		const workingSet = foldPlan(snapshot.text);
+		turnsStale = workingSet === lastWorkingSet ? turnsStale + 1 : 0;
+		lastWorkingSet = workingSet;
+		refresh(ctx);
+		if (turnsStale === 8 && goals(snapshot.text).some(g => g.status === "open" || g.status === "active")) {
+			// Pi queues context-only messages until tool results are appended at turn_end.
+			// This reaches the next model call in a long run without triggering another run.
+			pi.sendMessage({ customType: "pi-goals-upkeep", content: upkeep(state.plan!), display: false }, { triggerTurn: false });
+		}
+	});
+	pi.on("agent_end", (_e, ctx) => { refresh(ctx); if (!planWatcher && state.mode === "supervising") watchPlan(ctx); });
+	let proposedDraft = "";
+	let proposing = false;
+	pi.on("agent_settled", async (_e, ctx) => {
+		if (state.child || state.mode !== "planning" || !ctx.hasUI || proposing) return;
+		const text = planText();
+		const version = `${state.plan}:${digest(text)}`;
+		if (!goals(text).length || version === proposedDraft) return;
+		proposedDraft = version;
+		proposing = true;
+		try {
+			pi.sendMessage({ customType: "goal-plan-proposal", content: text, display: true }, { triggerTurn: false });
+			await ready(ctx, true);
+		} finally { proposing = false; }
+	});
+	// No context hook. Historical message arrays, native checkpoints and model selection are untouched.
+	pi.on("before_agent_start", (event, ctx) => {
+		if (state.mode === "chat") return;
+		const snapshot = readPlan();
+		if (snapshot.text === undefined) {
+			notice = true; // Retry resync on the next turn; do not consume a failed snapshot.
+			return { systemPrompt: `${event.systemPrompt}\n\n${state.child ? childPlanRole : ""}\n${snapshot.error}` };
+		}
+		const role = state.child ? childPlanRole : state.mode === "supervising"
+			? supervisor(WORKER, state.plan!, ctx.sessionManager.getSessionId())
+			: state.mode === "planning" ? planning(state.plan!) : state.mode === "paused" ? pausedRole : soloRole;
+		const content = notice ? planContext(state.child ? "worker" : state.mode, state.plan, snapshot.text)
+			: undefined;
+		if (content) turnsStale = 0;
+		notice = false;
+		return { systemPrompt: `${event.systemPrompt}\n\n${role}`, ...(content ? { message: { customType: "pi-goals-plan", content, display: false } } : {}) };
+	});
+	pi.on("tool_call", (event) => {
+		if (state.child || !["subagent", "subagent_resume"].includes(event.toolName)) return;
+		// Solo means this chat took over implementation: no concurrent writer may be delegated.
+		if (state.mode === "planning" || state.mode === "paused" || state.mode === "solo") return { block: true, reason: goalToolBlocked(state.mode) };
+		if (state.plan) { pendingLaunches++; state.workerStopped = false; workerRevision++; save(); }
+	});
+	pi.on("tool_result", (event) => {
+		if (state.child || !state.plan || !["subagent", "subagent_resume"].includes(event.toolName)) return;
+		pendingLaunches = Math.max(0, pendingLaunches - 1);
+		if (event.isError) return;
+		const details = event.details as { id?: string; sessionFile?: string } | undefined;
+		if (!details?.id || !details.sessionFile) return;
+		const record = { id: details.id, sessionFile: details.sessionFile };
+		if (state.worker?.sessionFile === record.sessionFile) state.worker = record;
+		else if (!state.worker) state.worker = record;
+		// Extra launches stay recorded as helpers; the implementation binding never moves silently.
+		else state.helpers = [...(state.helpers ?? []).filter((h) => h.sessionFile !== record.sessionFile), record];
+		state.workerStopped = false; workerRevision++; save();
+	});
 
 	pi.registerCommand("goals", {
-		description: `Plan mode: draft goals into ${PLAN_REL}, review, then work them. /goals <objective> | /goals clear | /goals judge <model>`,
+		description: "Goal plan actions: new, review, ready, status, stop, resume, solo, attach, model, exit",
+		getArgumentCompletions: (prefix) => ["new", "review", "ready", "status", "stop", "resume", "solo", "attach", "model", "exit", "help"].filter((verb) => verb.startsWith(prefix)).map((verb) => ({ value: verb, label: verb })),
 		handler: async (args, ctx) => {
-			const arg = args.trim();
-			if (arg === "clear") {
-				writePlan(ctx, "");
-				state = { ...state, isPlanMode: false };
-				persist();
-				updateWidget(ctx);
-				ctx.ui.notify(`Cleared ${PLAN_REL}.`, "info");
-				return;
-			}
-			if (arg.startsWith("judge")) {
-				const ref = arg.slice("judge".length).trim();
-				state = { ...state, judgeModel: ref || null };
-				persist();
-				ctx.ui.notify(ref ? `Sign-off judge model set to ${ref}` : "Sign-off judge reset to the session model", "info");
-				return;
-			}
-			state = { ...state, isPlanMode: true };
-			persist();
-			updateWidget(ctx);
-			// The drafting rules are sent ONCE, with the seed. v2 re-injected them every turn, which is
-			// why plan mode read as never-ending: every reply re-armed it. They come back only on a
-			// resync (session start / compaction), when the model has genuinely lost them.
-			const seed = arg
-				? `We're in plan mode. Objective: ${arg}\n\n${planDrafting}\n\nWrite the plan to ${planPath(ctx)}.`
-				: `We're in plan mode. Tell me what you want to plan.\n\n${planDrafting}\n\nWrite the plan to ${planPath(ctx)}.`;
-			pi.sendUserMessage(seed, { deliverAs: "followUp" });
-		},
-	});
-
-	// --- hooks --------------------------------------------------------------------------------------
-
-	/** What this LLM call should carry, if anything: a one-shot resync, or a staleness reminder. */
-	function dueInjection(ctx: ExtensionContext, plan: string): string | null {
-		const drainResync = (): string | null => {
-			const why = resyncReason;
-			resyncReason = null;
-			return why;
-		};
-		if (state.isPlanMode) {
-			const why = drainResync();
-			return why ? `<system-reminder>\n${why} You are still in plan mode.\n\n${planDrafting}\n\nWrite the plan to ${planPath(ctx)}.\n</system-reminder>` : null;
-		}
-		if (!plan.trim()) return null;
-		const why = drainResync();
-		if (why) return resync(plan, PLAN_REL, why);
-		if (turnsStale < STALE_TURNS) return null;
-		const goals = scanGoals(plan);
-		if (goals.length === 0) {
-			// Non-empty plan but no recognizable goal line: the harness would go silently inert (no
-			// widget, no injection, no reminders). Say so instead -- cooperative but confused.
-			return `<system-reminder>\n${PLAN_REL} exists but has no goal line pi-goals recognizes. A goal is a checkbox list line starting "goal:", e.g. "1. [ ] goal: <imperative>" ([ ] open, [/] active, [x] done, [-] cancelled). Reformat it if it's meant to be the plan.\n</system-reminder>`;
-		}
-		if (!goals.some((g) => g.status === "active" || g.status === "open")) return null;
-		return reminder(foldPlan(plan), PLAN_REL);
-	}
-
-	// The one injection point: a transient user message on this LLM call only, never persisted. So
-	// there are no stale copies to strip, and an untouched turn costs nothing.
-	pi.on("context", async (event, ctx) => {
-		const text = dueInjection(ctx, readPlan(ctx));
-		if (!text) return;
-		turnsStale = 0;
-		return { messages: [...event.messages, { role: "user" as const, content: [{ type: "text" as const, text }], timestamp: Date.now() }] };
-	});
-
-	// The staleness clock: editing the plan resets it, the way a task tool call resets pi-tasks'.
-	pi.on("turn_end", async (_event, ctx) => {
-		const plan = readPlan(ctx);
-		if (plan === lastSeenPlan) {
-			turnsStale++;
-			return;
-		}
-		lastSeenPlan = plan;
-		turnsStale = 0;
-		updateWidget(ctx);
-	});
-
-	// A compaction is exactly when the settled context is gone, so push the whole file back once.
-	pi.on("session_compact", async () => {
-		resyncReason = "The session was just compacted.";
-	});
-
-	// Plan mode gate: block edit/write except on plan.md itself. bash stays open (guide, not gate).
-	pi.on("tool_call", async (event, ctx) => {
-		if (!state.isPlanMode) return;
-		if (PLAN_MODE_BLOCKED_TOOLS.includes(event.toolName)) {
-			const target = (event.input as { path?: string }).path;
-			if (target && resolve(ctx.cwd, target) === resolve(planPath(ctx))) return;
-			return { block: true, reason: `Plan mode is read-only: only ${PLAN_REL} may be written while drafting. Agree the goals first, then choose Ready.` };
-		}
-	});
-
-	// After a plan-mode turn: if goals were drafted, offer Ready. The human reads the file and says
-	// go, edits it in $EDITOR, or keeps talking to revise it (menu shape borrowed from pi-plan).
-	pi.on("agent_end", async (_event, ctx) => {
-		if (!state.isPlanMode || !ctx.hasUI) return;
-		while (scanGoals(readPlan(ctx)).length > 0) {
-			const choice = await ctx.ui.select(`Plan drafted in ${PLAN_REL}. Ready?`, [
-				"Ready — start working the plan",
-				"Open in $EDITOR — edit it myself",
-				"Keep planning (reply to revise)",
-			]);
-			if (choice?.startsWith("Open")) {
-				spawnSync(process.env.EDITOR || process.env.VISUAL || "vi", [planPath(ctx)], { stdio: "inherit" });
-				continue;
-			}
-			if (!choice?.startsWith("Ready")) return;
-			state = { ...state, isPlanMode: false };
-			persist();
-			updateWidget(ctx);
-			pi.sendUserMessage(
-				`Work the goals in ${planPath(ctx)}. Pick an open goal, mark it active ([/]), work its subtasks, and when its discriminator is satisfied fill its evidence: list, then call CompleteGoal with the goal's text. Keep the plan file current as you go.`,
-				{ deliverAs: "followUp" },
-			);
-			return;
-		}
-	});
-
-	pi.on("session_start", async (_event, ctx) => {
-		// v1 wrote .pi/goals.md; v2 reads .pi/plan.md. Rename so old goals aren't silently invisible
-		// (dogfood finding). Claude: one-time migration, delete once v1 files are gone from the wild.
-		const v1Path = join(ctx.cwd, ".pi", "goals.md");
-		if (existsSync(v1Path) && !existsSync(planPath(ctx))) {
-			renameSync(v1Path, planPath(ctx));
-			ctx.ui.notify(`Renamed .pi/goals.md -> ${PLAN_REL} (v2 filename).`, "info");
-		}
-		const last = ctx.sessionManager
-			.getEntries()
-			.filter((e: { type?: string; customType?: string }) => e.type === "custom" && e.customType === STATE)
-			.pop() as { data?: PlanState } | undefined;
-		if (last?.data) state = { ...state, ...last.data };
-		lastSeenPlan = readPlan(ctx);
-		resyncReason = "New session.";
-		updateWidget(ctx);
-	});
-
-	// --- the one blessed tool: CompleteGoal ---------------------------------------------------------
-
-	pi.registerTool({
-		name: "CompleteGoal",
-		label: "Goal signoff",
-		description: completeGoalDescription,
-		parameters: Type.Object({
-			goal: Type.String({ description: completeGoalParamDescription }),
-		}),
-		async execute(_id, params, signal, onUpdate, ctx) {
-			const plan = readPlan(ctx);
-			if (!plan.trim()) return result(`No plan file at ${PLAN_REL}.`, true);
-
-			const judgeModel = state.judgeModel ?? (ctx.model ? `${ctx.model.provider}/${ctx.model.id}` : null);
-			onUpdate?.({ content: [{ type: "text", text: `Read-only judge (${judgeModel ?? "pi default"}) inspecting: ${params.goal}` }], details: {} });
-			// decideSignOff runs the judge and derives the outcome + the one log line. judgeModel is never
-			// checked pre-emptively: null just means pi's configured default runs (buildJudgeArgs omits
-			// --model), so accepted_inconclusive always means "the judge ran but failed", never "no model".
-			let judgeRaw: JudgeResult | null = null;
-			const outcome = await decideSignOff({ goal: params.goal, plan, judgeModel }, signal, async (task) => {
-				judgeRaw = await runJudge(task, judgeModel, ctx.cwd, signal);
-				return judgeRaw;
-			});
-			// Persist the judge's full transcript so "did the judge really re-run verify?" is answerable
-			// after the fact (dogfood finding: with only the one log line, an accept is unauditable).
-			let transcriptNote = "";
-			if (judgeRaw !== null) {
-				const raw: JudgeResult = judgeRaw;
-				mkdirSync(join(ctx.cwd, ".pi", "judge"), { recursive: true });
-				const rel = `.pi/judge/${stamp().replace(/[: ]/g, "-")}.md`;
-				writeFileSync(join(ctx.cwd, rel), `goal: ${params.goal}\nmodel: ${judgeModel ?? "pi default"}\nerror: ${raw.error ?? "none"}\n\n${raw.output}\n`);
-				transcriptNote = ` (${rel})`;
-			}
-			if (outcome.logEntry) {
-				// Sign-off write: tick the goal [x] (exact-subject match; dogfood showed agent bookkeeping
-				// is the drift point) and append the audit log line, one write. On wording drift the tick
-				// falls to the agent and the result says so -- both paths are explicit, never silent.
-				let updated = readPlan(ctx);
-				let tickNote = "";
-				if (outcome.logEntry.startsWith("signed off")) {
-					const ticked = tickGoal(updated, params.goal);
-					updated = ticked ?? updated;
-					tickNote = ticked
-						? `\n\nGoal ticked [x] in ${PLAN_REL}.`
-						: `\n\nNo exact goal line matched your wording -- tick it [x] in ${PLAN_REL} yourself.`;
+			try {
+				if (state.child) { ctx.ui.notify("This is the delegated worker. Goal approval belongs to its parent.", "info"); return; }
+				let command = args.trim();
+				if (!command) {
+					const actions = ["status — Show current plan", "new — New plan", "attach — Open an existing plan", "review — Review current plan", "ready — Approve draft", "stop — Pause work", "resume — Continue paused work", "solo — Work in this session", "model — Set worker model", "exit — Leave goal mode", "help — Show commands"];
+					const before = generation;
+					const choice = await ctx.ui.select("Goal plan actions", actions);
+					if (!choice || before !== generation) return;
+					command = choice.split(" — ")[0];
+					if (["attach", "model"].includes(command)) {
+						const value = await ctx.ui.editor(command === "attach" ? "Plan path (optional: solo)" : "Worker model (provider/model)", "");
+						if (!value?.trim() || before !== generation) return;
+						command += ` ${value.trim()}`;
+					}
 				}
-				writePlan(ctx, appendLog(updated, `${stamp()} ${outcome.logEntry}${transcriptNote}`));
-				updateWidget(ctx);
-				return result(outcome.resultText + tickNote, outcome.isError);
-			}
-			return result(outcome.resultText, outcome.isError);
+				if (command === "help") { ctx.ui.notify(help, "info"); return; }
+				if (command === "status") {
+					refresh(ctx);
+					ctx.ui.notify([
+						`Mode: ${state.mode}`,
+						`Plan: ${state.plan ?? "none"}`,
+						`Preferred worker model (plan): ${notedPlanValue("preferred worker model") ?? "not stated; use /goals model <model>"}`,
+						`Recorded worker session: ${state.worker?.sessionFile ?? "not recorded"}`,
+						`Helper subagent sessions: ${state.helpers.length} recorded (liveness via /subagents)`,
+						notedPlanValue("worker session") ? `Worker session noted in plan: ${notedPlanValue("worker session")}` : "",
+						`Hourly check-in: schedule_prompt job ${JSON.stringify(`goals-${ctx.sessionManager.getSessionId()}`)} (list/remove via schedule_prompt; plan-change reviews are the plan-watcher event hook)`,
+						"Liveness is owned by edxeth; inspect /subagents.",
+					].filter(Boolean).join("\n"), "info");
+					return;
+				}
+				if (command === "review" && state.mode === "supervising") { notice = true; send(manualReview(state.plan ?? "")); return; }
+				if (command === "review" || command === "ready") { await ready(ctx, command === "review"); return; }
+				if (command === "model" || command.startsWith("model ")) {
+					if (!state.plan || !goals(planText()).length) { ctx.ui.notify("Register a goal plan first.", "warning"); return; }
+					const ref = command.slice("model".length).trim();
+					if (!ref) { ctx.ui.notify("Use /goals model <provider/model>; no preference changed.", "info"); return; }
+					const lines = planText().split("\n");
+					const pref = `- preferred worker model: ${ref || "(none specified)"}`;
+					const found = lines.findIndex((line) => /^-\s*preferred worker model:/i.test(line));
+					if (found >= 0) lines[found] = pref;
+					else { const title = lines.findIndex((line) => /^#\s/.test(line)); lines.splice(title >= 0 ? title + 1 : 0, 0, pref); }
+					writeFileSync(state.plan, lines.join("\n"));
+					planHash = digest(planViews(planText()).notify);
+					refresh(ctx);
+					ctx.ui.notify(ref ? `Preferred worker model set to ${ref} in plan preferences. The supervisor selects it at launch and verifies the resolved model; the worker pane's own model is chosen with /model in that pane.` : "Preferred worker model cleared.", "info");
+					return;
+				}
+				if (command === "attach" || command.startsWith("attach ")) {
+					const rest = command.slice("attach".length).trim();
+					const [raw, kind, extra] = rest.split(/\s+/);
+					const solo = kind === "solo";
+					if (extra || (kind && !solo)) { ctx.ui.notify("Use /goals attach <path-to-plan.md> [solo].", "warning"); return; }
+					if (!raw) { ctx.ui.notify("Use /goals attach <path-to-plan.md> [solo].", "info"); return; }
+					const target = isAbsolute(raw) ? raw : resolve(ctx.cwd, raw);
+					let text: string;
+					try { text = readFileSync(target, "utf8"); } catch { ctx.ui.notify(`Cannot read plan at ${target}.`, "error"); return; }
+					if (!goals(text).length) { ctx.ui.notify(`${target} has no '- [ ] goal:' lines; attach a judgeable plan.`, "warning"); return; }
+					if (!solo && ((state.worker && !state.workerStopped) || state.mode === "supervising")) { ctx.ui.notify("Exit and resolve the existing worker before replacing the plan. The current plan is preserved.", "warning"); return; }
+					const noted = /^-\s*worker session:\s*(\S+)/im.exec(foldPlan(text))?.[1];
+					if (!(await confirmOwnership(ctx, target, text, solo))) return;
+					const retained = target === state.plan ? state.signoffs : {};
+					const worker = noted ? { sessionFile: resolve(ctx.cwd, noted) } : state.workerStopped ? state.worker : undefined;
+					state = { mode: solo ? "solo" : "planning", plan: target, signoffs: retained, worker, helpers: [], workerStopped: solo || (!noted && state.workerStopped) };
+					generation++; notice = true; save(); refresh(ctx); watchPlan(ctx);
+					if (solo) enterSolo(ctx);
+					else send(attachNotice(target, false, noted));
+					return;
+				}
+				if (command === "stop" || command === "exit") {
+					if (state.mode === "planning") {
+						if (command === "stop") { ctx.ui.notify("A draft cannot pause; use /goals exit to leave planning with the draft preserved.", "warning"); return; }
+						// Planning exit must not get the model trapped re-planning or lose the draft.
+						state.mode = "chat"; generation++; notice = true; save(); refresh(ctx); watchPlan(ctx);
+						ctx.ui.notify(`Planning exited; draft preserved at ${state.plan}. No implementation was approved or started. Reconnect with /goals attach ${state.plan}.`, "info");
+						return;
+					}
+					state.mode = command === "stop" ? "paused" : "chat"; generation++; notice = true; save(); refresh(ctx); watchPlan(ctx);
+					send(`${removeGoalSchedule(ctx.sessionManager.getSessionId())}\n\n${pauseExitNotice(state.worker, command === "exit")}`, Boolean(state.worker) || hasScheduleTool());
+					return;
+				}
+				if (command === "resume") {
+					if (state.mode !== "paused" || !state.plan) { ctx.ui.notify("Only a paused approved plan can resume. A draft needs Ready.", "warning"); return; }
+					if (!compatible()) { ctx.ui.notify("edxeth tools unavailable; plan remains paused.", "error"); return; }
+					state.mode = "supervising"; generation++; notice = true; save(); refresh(ctx); watchPlan(ctx);
+					send(`${checkIn(ctx)}\n\n${resumeNotice(WORKER, state.plan, state.worker)}`);
+					return;
+				}
+				if (command === "solo") {
+					if (!state.plan || !goals(planText()).length) { ctx.ui.notify("Register a goal plan first.", "warning"); return; }
+					if (!(await confirmOwnership(ctx, state.plan, planText()))) return;
+					enterSolo(ctx);
+					return;
+				}
+				if (command !== "new" && !command.startsWith("new ")) { ctx.ui.notify(`Unknown or incomplete command. ${help}`, "warning"); return; }
+				const objective = command.slice(4).trim();
+				if ((state.worker && !state.workerStopped) || state.mode === "supervising") { ctx.ui.notify("Exit and resolve the existing worker before replacing the plan. The current plan is preserved.", "warning"); return; }
+				const path = join(ctx.cwd, ".pi", "plan", `${ctx.sessionManager.getSessionId()}-main.md`);
+				mkdirSync(dirname(path), { recursive: true });
+				// Never overwrite an earlier plan at this session path; the model can revise it after inspection.
+				try { writeFileSync(path, planDocument(objective), { flag: "wx" }); } catch (error) { if ((error as NodeJS.ErrnoException).code !== "EEXIST") throw error; }
+				state = { mode: "planning", plan: path, signoffs: {}, worker: state.worker, helpers: state.helpers, workerStopped: state.workerStopped }; generation++; notice = true; save(); refresh(ctx); watchPlan(ctx);
+				send(planningSeed(objective, path));
+			} catch (error) { ctx.ui.notify(String(error), "error"); }
+		},
+	});
+	pi.registerTool({
+		name: "AttachGoalPlan", label: "Attach delegated plan", description: attachGoalPlanDescription,
+		parameters: Type.Object({ path: Type.String() }),
+		async execute(_id, params, _signal, _update, ctx) {
+			if (!state.child) return result(messages.childAttachOnly);
+			try {
+				if (!isAbsolute(params.path) || !goals(readFileSync(params.path, "utf8")).length) return result(messages.invalidAttachment);
+			} catch { return result(messages.invalidAttachment); }
+			state.plan = params.path; generation++; notice = true; save(); refresh(ctx);
+			return result(childPlanAttached(params.path));
+		},
+	});
+	pi.registerTool({
+		name: "CompleteGoal", label: "Review goal evidence",
+		description: completeGoalDescription,
+		parameters: Type.Object({ goal: Type.String(), evidence: Type.Array(Type.String(), { minItems: 1 }), observation: Type.String({ minLength: 1 }) }),
+		async execute(_id, params, signal, _update, ctx) {
+			if (state.child || !["supervising", "solo"].includes(state.mode)) return result(messages.completionUnavailable);
+			if (signal?.aborted) return result(messages.cancelled);
+			const snapshot = readPlan();
+			if (snapshot.text === undefined) return result(snapshot.error!);
+			const text = snapshot.text;
+			const matches = goals(text).filter((g) => g.status !== "cancelled" && key(g.subject) === key(params.goal));
+			if (matches.length !== 1 || !state.plan) return result(messages.uniqueGoal);
+			const evidence = params.evidence.map((file) => isAbsolute(file) ? file : resolve(ctx.cwd, file));
+			try { for (const file of evidence) if (!readFileSync(file).length) throw new Error(emptyEvidence(file)); }
+			catch (error) { return result(evidenceUnavailable(error)); }
+			const lines = text.split("\n");
+			lines[matches[0].index] = lines[matches[0].index].replace(/\[[ xX/-]\]/, "[x]");
+			let log = lines.findIndex(line => /^##\s+Log\s*$/i.test(line));
+			if (log === -1) { lines.push("", "## Log"); log = lines.length - 1; }
+			lines.splice(log + 1, 0, "", completionLog(params.goal, params.observation, evidence, state.mode === "solo"));
+			writeFileSync(state.plan, `${lines.join("\n").trimEnd()}\n`);
+			state.signoffs[key(matches[0].subject)] = { evidence, observation: params.observation };
+			planHash = digest(planViews(planText()).notify);
+			save(); refresh(ctx);
+			const remaining = goals(planText()).some((goal) => goal.status !== "cancelled" && (goal.status !== "done" || !state.signoffs[key(goal.subject)]));
+			return result(completionResult(matches[0].subject, ctx.sessionManager.getSessionId(), remaining, state.mode === "solo"));
 		},
 	});
 }
-
-// --- helpers (module scope) --------------------------------------------------------------------
-
-function result(text: string, isError = false) {
-	return { content: [{ type: "text" as const, text }], details: {}, isError };
-}
-
-/** Local time, not UTC: agents freehand-stamp their manual ## Log lines from the local clock they
- *  see, so a UTC tool stamp made the trail read as two different afternoons (dogfood finding). */
-function stamp(): string {
-	const d = new Date();
-	const p = (n: number) => String(n).padStart(2, "0");
-	return `${d.getFullYear()}-${p(d.getMonth() + 1)}-${p(d.getDate())} ${p(d.getHours())}:${p(d.getMinutes())}`;
-}
-
-function oneLine(s: string): string {
-	return s.replace(/\s+/g, " ").trim().slice(0, 200);
-}
-
-/** A judge run's result: stdout output, plus an error string when the subprocess failed/timed out. */
-export interface JudgeResult {
-	output: string;
-	error?: string;
-}
-
-/** Inputs to a sign-off decision. judgeModel is null when no explicit/session model is set. */
-export interface SignOffInput {
-	goal: string;
-	plan: string;
-	judgeModel: string | null;
-}
-
-/** The outcome of a sign-off: the reply text, whether it's a hard error, and the one ## Log line to
- *  append (null when nothing should be written, e.g. aborted before any verdict). */
-export interface SignOffOutcome {
-	resultText: string;
-	isError: boolean;
-	logEntry: string | null;
-}
-
-/** Run the judge and decide accept / reject / accepted_inconclusive. Pure aside from the injected
- *  judge runner, so the unit test can lock the fail-forward invariant: judgeModel is NEVER checked
- *  here, so a null model still reaches runJudge (pi's configured default runs it), and the only
- *  producers of accepted_inconclusive are the judge-error and no-VERDICT paths -- i.e. "the judge
- *  ran but failed", never "no model". The execute() wrapper does the plan-file write + widget.
- *  Exported for the unit test that locks this invariant. */
-export async function decideSignOff(
-	input: SignOffInput,
-	signal: AbortSignal | undefined,
-	runJudgeFn: (task: string) => Promise<JudgeResult>,
-): Promise<SignOffOutcome> {
-	const task = judgeUser({ goal: input.goal, plan: input.plan, planPath: PLAN_REL });
-	const judge = await runJudgeFn(task);
-
-	if (signal?.aborted) return { resultText: "Sign-off aborted.", isError: true, logEntry: null };
-
-	// Judge ran but failed/errored/timed out: fail forward, say so in the log.
-	if (judge.error) {
-		const partial = judge.output ? `\n\npartial judge output:\n${judge.output}` : "";
-		return {
-			resultText: `Judge ran but failed (${judge.error}). Accepted inconclusive — logged.${partial}`,
-			isError: false,
-			logEntry: `signed off "${input.goal}" (judge inconclusive: ran but failed: ${oneLine(judge.error)})`,
-		};
-	}
-
-	const verdictLine = judge.output.split("\n").find((l) => /^\s*VERDICT\s*:/i.test(l)) ?? "";
-	const verdict = /^\s*VERDICT\s*:\s*(accept|reject)\s*$/i.exec(verdictLine)?.[1]?.toLowerCase();
-	const reasoning = judge.output.length > 2000 ? `...\n${judge.output.slice(-2000)}` : judge.output;
-
-	if (verdict === "accept") {
-		return {
-			resultText: `Sign-off ACCEPTED (log line appended).\n\n--- judge ---\n${reasoning}`,
-			isError: false,
-			logEntry: `signed off "${input.goal}" (judge accept)`,
-		};
-	}
-	if (verdict === "reject") {
-		const missing = judge.output.match(/missing\s*:\s*([\s\S]*)$/i)?.[1].trim() || judge.output.slice(-500);
-		return {
-			resultText: `Sign-off REJECTED. Missing:\n${missing}\n\n--- judge ---\n${reasoning}`,
-			isError: true,
-			logEntry: `reject "${input.goal}": ${oneLine(missing)}`,
-		};
-	}
-	// No VERDICT line: same fail-forward as a judge error -- the judge ran but didn't answer.
-	return {
-		resultText: `Judge returned no VERDICT line. Accepted inconclusive — logged.\n\n--- judge ---\n${reasoning || "(no output)"}`,
-		isError: false,
-		logEntry: `signed off "${input.goal}" (judge inconclusive: no VERDICT line)`,
-	};
-}
-
-/** Tick the goal line whose subject exactly matches `goal` (trimmed, case-insensitive) to [x].
- *  Null when there is no unique exact match (wording drift / duplicates) -- the caller then asks the
- *  agent to tick it itself. Reuses GOAL_LINE; deliberately NOT fuzzy, that's the judge's job. */
-export function tickGoal(plan: string, goal: string): string | null {
-	const lines = plan.split("\n");
-	const want = goal.trim().toLowerCase();
-	const hits = lines.flatMap((l, i) => (GOAL_LINE.exec(l)?.[2].trim().toLowerCase() === want ? [i] : []));
-	if (hits.length !== 1) return null;
-	lines[hits[0]] = lines[hits[0]].replace(/\[[ xX/-]\]/, "[x]");
-	return lines.join("\n");
-}
-
-/** Append one line under ## Log (creating the section at EOF if absent). */
-export function appendLog(text: string, entry: string): string {
-	const lines = text.split("\n");
-	const line = `- ${entry}`;
-	const header = lines.findIndex((l) => /^##\s+Log\s*$/i.test(l));
-	if (header === -1) return `${text.replace(/\n+$/, "")}\n\n## Log\n${line}\n`;
-	let insertAt = header + 1;
-	for (let i = header + 1; i < lines.length; i++) {
-		if (/^#{1,6}\s/.test(lines[i])) break;
-		if (/^\s*-\s+/.test(lines[i])) insertAt = i + 1;
-	}
-	lines.splice(insertAt, 0, line);
-	return lines.join("\n");
-}
-
-/** Build the pi argv for the read-only judge. `--model` is omitted when no explicit/session model is
- *  set, so pi falls back to its configured default — the judge always runs. `--no-extensions` keeps
- *  the judge minimal and immune to a broken third-party extension taking down every sign-off.
- *  Exported for the unit test that locks these invariants. */
-export function buildJudgeArgs(judgeModel: string | null): string[] {
-	const args = ["-p", "--no-session", "--no-extensions"];
-	if (judgeModel) args.push("--model", judgeModel);
-	args.push("--tools", JUDGE_TOOLS.join(","), "--exclude-tools", JUDGE_BLOCKED_TOOLS.join(","), "--append-system-prompt", judgeSystem);
-	return args;
-}
-
-/** Locate the pi binary the same way the oracle extension does, so spawning works under bun or node. */
-function getPiInvocation(args: string[]): { command: string; args: string[] } {
-	const script = process.argv[1];
-	if (script && !script.startsWith("/$bunfs/root/") && existsSync(script)) return { command: process.execPath, args: [script, ...args] };
-	const execName = basename(process.execPath).toLowerCase();
-	if (!/^(node|bun)(\.exe)?$/.test(execName)) return { command: process.execPath, args };
-	return { command: "pi", args };
-}
-
-/** Spawn the read-only judge subprocess (plain `pi -p`: stdout is the final response text). */
-async function runJudge(
-	task: string,
-	judgeModel: string | null,
-	cwd: string,
-	signal: AbortSignal | undefined,
-): Promise<JudgeResult> {
-	const args = buildJudgeArgs(judgeModel);
-	args.push(task);
-	const inv = getPiInvocation(args);
-	// Runs in-place against this checkout; pi --no-session does not clone into the parent
-	// (proven by scripts/check-judge-footprint.sh).
-	return new Promise((resolvePromise) => {
-		let settled = false;
-		let stdout = "";
-		let stderr = "";
-		const done = (r: { output: string; error?: string }) => {
-			if (!settled) {
-				settled = true;
-				clearTimeout(timer);
-				resolvePromise(r);
-			}
-		};
-		const proc = spawn(inv.command, inv.args, { cwd, shell: false, stdio: ["ignore", "pipe", "pipe"], signal });
-		const timer = setTimeout(() => {
-			proc.kill();
-			done({ output: stdout.trim(), error: `judge timed out after ${JUDGE_TIMEOUT_MS / 1000}s` });
-		}, JUDGE_TIMEOUT_MS);
-		proc.stdout?.on("data", (d) => {
-			stdout += d.toString();
-		});
-		proc.stderr?.on("data", (d) => {
-			stderr += d.toString();
-		});
-		proc.on("close", (code) => {
-			if ((code ?? 0) !== 0) done({ output: stdout.trim(), error: stderr.trim() || `judge subprocess exited ${code ?? 1}` });
-			else done({ output: stdout.trim() });
-		});
-		proc.on("error", (e) => done({ output: stdout.trim(), error: `judge subprocess failed: ${e.message}` }));
-	});
-}

src/plan-view.ts

@@ -0,0 +1,33 @@
+// Pi/OpenAI: Preserve plan wording; omit history and, in the short view, task/evidence details.
+// The notify view governs plan-change events: goals, tasks, evidence and inferences are
+// content worth a supervisor review; worker identity bookkeeping is not (field report,
+// LUCID3 supervisor 2026-09-10: two identical review events for a session-path edit).
+export function planViews(plan: string): { short: string; notify: string; long: string } {
+	const long = plan.split(/^#{1,6}\s+(?:Log|Appendix|Appendices|Appendixes|Interview|Learnings|Papercuts)\b.*$/mi)[0].trim();
+	const identity = /^-\s*(?:active worker|worker session|worker intercom session):/i;
+	const notify = long.split("\n").filter((line) => !identity.test(line)).join("\n").trim();
+	const kept: string[] = [];
+	let omittedIndent: number | null = null;
+	let omittedHeading: number | null = null;
+	for (const line of long.split("\n")) {
+		// Pi/OpenAI: Worker identity bookkeeping is not a change to agreed requirements.
+		if (/^-\s*(?:active worker|worker session|worker intercom session):/i.test(line)) continue;
+		const heading = /^(#{1,6})\s+(.+)$/.exec(line);
+		if (heading) {
+			if (omittedHeading !== null && heading[1].length <= omittedHeading) omittedHeading = null;
+			if (/^(?:Tasks?|Task list|Subtasks?|Evidence)\b/i.test(heading[2])) omittedHeading = heading[1].length;
+		}
+		if (omittedHeading !== null) continue;
+		const indent = line.match(/^\s*/)?.[0].length ?? 0;
+		if (omittedIndent !== null) {
+			if (!line.trim() || indent > omittedIndent) continue;
+			omittedIndent = null;
+		}
+		if (/^\s*[-*]\s+(?:tasks?|subtasks?|evidence):/i.test(line) || /^\s*(?:\d+[.)]|[-*])\s+\[[ x/~-]\]\s+(?!goal:)/i.test(line)) {
+			omittedIndent = indent;
+			continue;
+		}
+		kept.push(line);
+	}
+	return { short: kept.join("\n").trim(), notify, long };
+}

src/plan.ts

@@ -0,0 +1,8 @@
+// Shared plan syntax: only the section above the Log contains current goals.
+export const GOAL_LINE = /^\s*(?:\d+\.|[-*])\s*\[([ xX/-])\]\s*goal:\s*(.*)$/i;
+export const FOLD_LINE = /^##\s+Log\s*$/im;
+
+export function foldPlan(plan: string): string {
+	const match = FOLD_LINE.exec(plan);
+	return (match ? plan.slice(0, match.index) : plan).trimEnd();
+}

src/prompts.ts

@@ -1,78 +1,79 @@
-/**
- * pi-goals v2 — all model-facing text, in flow order.
- *
- * Design: plan.md is for LLMs and the human, not for TypeScript. There is no parser and no schema;
- * the skeleton below is a convention the drafting prompt teaches, the working agent maintains with
- * its normal Edit tool, and the judge reads natively. The harness does three things for a
- * cooperative-but-confused model: memory (a transient re-send of the plan when it goes stale),
- * format guidance (the skeleton), and fresh eyes (the read-only judge in CompleteGoal).
- *
- * THE FOLD: everything above "## Log" is the working set (title, user voice, goals,
- * discriminators) and is what gets re-sent on the reminder cadence. Everything below it (Log,
- * Learnings, Appendix) is durable memory: unlimited, read on demand, and re-sent in full only at
- * session start and after a compaction, which is where the settled context is actually needed.
- *
- * Flow:
- *   SETUP (plan mode)     1. planDrafting   — draft goals into plan.md (read-only phase), sent once
- *   EXEC, on cadence      2. reminder       — the folded plan + upkeep nudge when plan.md went stale
- *   EXEC, after compact   3. resync         — the WHOLE file back, once
- *   SIGN-OFF, agent-side  4. completeGoal*  — the one blessed tool's description
- *   SIGN-OFF, judge-side  5. judgeSystem/judgeUser — the one rigorous check
- *
- * The goal's test is the DISCRIMINATOR: the concrete observation that tells real success from the
- * named subtle failure mode. Evidence is empty at planning and filled at sign-off.
- */
-
-/* ─────────────────────────────────────────────────────────────────────────
- * 1. planDrafting — SETUP, plan mode (read-only: edit/write blocked except plan.md)
- * ──────────────────────────────────────────────────────────────────────── */
+// Pi/OpenAI: Planning, approval, supervision, reminders, completion and recovery.
 export const planDrafting = `\
-You are in plan mode. The objective may arrive through conversation, not as one up-front command.
-Explore the repository read-only first: resolve discoverable facts by looking them up, and only ask
-the human when the answer is a genuine intent or preference choice. Do not write or run code in this
-phase (edit/write are blocked except for the plan file; don't mutate state via bash either). When
-the objective is clear, draft the plan file and present it.
+You are in plan mode. Help the user express what they want this project to achieve in a short judgeable plan. Seek to understand their underlying goals, infer ordinary details, and use their applicable AGENTS.md instructions, relevant skills, and project context to interpret the request correctly. Do not silently substitute your own goals or expand the agreed scope.
 
-How this mode ends: after each of your turns the human gets a menu (Ready / open in $EDITOR / keep
-planning). Plan mode ends when they pick Ready. So close every draft with one line -- the plan is
-final, pick Ready to start or reply to revise -- and do not redraft in silence. When a new
-requirement arrives, fold it in, say what changed, and say the plan is final again. Detail that
-doesn't change a goal or a discriminator belongs in the appendix, not in the goals.
+1. Reduce technical uncertainty first. Use read-only repository tools or web search when either can
+resolve a fact. Do not write or run code in this phase (edit/write are blocked except for the plan
+file; don't mutate state via bash either).
+2. Before you draft a goal, identify its object, observable result, scope, and any decision that the
+human would need to approve later. Briefly reframe the request in your own words to check comprehension
+and make your understanding visible: the intended outcome, boundary, and success check. Invite correction,
+but do not require confirmation when these are already clear. Ask questions that expose differences
+between your understanding and the user's that would otherwise stay hidden. Probe consequential
+assumptions, challenge inconsistencies, and follow up where an answer exposes a gap. Do not use a question quota or ask the human
+to approve ordinary implementation details. Inspect files or search the web before asking when either
+can answer a fact. If the human does not answer a question, record that
+point as unknown; do not silently replace it with an inference or turn it into a new blocking decision.
+Do not present the review menu with a placeholder goal such as "work out the thing", "improve it", or
+"investigate".
+3. Use questions to clarify and narrow the goal, test your assumptions, and bring your understanding
+into agreement with the user's. Respect their limited time: batch independent high-impact questions
+in one short round, where the answer materially reduces uncertainty
+while discovering the right plan. Each question must be short and self-contained: state the relevant
+context, use the human's language and ASD-STE100
+Simple Technical English, and give a recommended answer. Record each answer, or the unanswered
+unknown, in ## Interview. Draft goals and present Ready when the requested work is otherwise executable.
+Only withhold Ready for an unanswered choice that changes scope, spending, or the user-visible result.
+4. State the user-visible result before the goals: one concrete sentence naming what the human will
+inspect when this plan is done. Take it from the original request, not from your implementation plan.
+Every requested artifact and action must survive into this sentence. An agent-inferred constraint may
+not replace, defer, or contradict it; ask the human if an inference would change the result.
+5. When every goal has an object, observable result, settled scope, and required approval, draft the
+plan file and present it. It should be safe to work overnight and present the requested outcome.
+
+How this mode ends: after each settled draft the human gets a menu (Ready / Refine / Edit / Cancel).
+Plan mode ends only when they pick Ready. Refine collects short revision notes. Edit opens the full
+plan. When a new requirement arrives, fold it in, say what changed, and present the plan again.
+Detail that doesn't change a goal or a discriminator belongs in the appendix, not in the goals.
 
 Right-size it:
-- Default to ONE goal. Add another only when it's a genuinely separate checkpoint that can pass or
-  fail on its own. Most objectives are 1-2 goals.
+- One goal per distinct judgeable outcome. Group related goals when it helps judge them together
+  and readability. The count flows from the outcomes.
+- Describe outcomes in qualitative terms the supervisor and user can discriminate.
+	- Use the users language or more precise don't transform "MV" into "knob" as it looses precision and is overloaded
+	- Don't invent metrics or thresholds for problems you haven't explored yet - the supervisor should know it when it sees the outcome.
+  	- Quantitative gates are fine only when you are certain they survive contact with reality.
 - Subtasks are the steps inside a goal; add them when a goal has 3+ distinct steps, skip otherwise.
-- Don't invent goals to look thorough. When in doubt, merge.
-- Everything above "## Log" is the part the model carries while it works. Keep it under 50 lines,
-  reviewable in one pass. Everything below "## Log" is unlimited.
+- Two goals that share one discriminator are one goal. Merge them.
+- Keep the goal subject short. Put its important scope, failure modes, discriminator, tasks, and evidence in the indented block beneath it. The supervisor reads the whole block and the whole plan.
+- Keep the working set under 50 lines, excluding ## User voice. ## User voice has no line limit: quote
+  the human fully rather than shorten or paraphrase them. Everything below "## Log" is unlimited.
 
-Style: ASD-STE100 Simplified Technical English. Active voice, one idea per sentence, common words,
-the same word for the same thing, and define a new term at first use. This covers the context
-paragraph and the appendix too, not just the checklist. No all-caps headers and no bold spam; the
-checklist is already the structure.
+Style: Make it easy for a busy and forgetfull user to review. Use ASD-STE100 Simplified Technical English. Use active voice, one idea per sentence, common words,
+the same word for the same thing, and define a new terms at first use. Use redundant context for skim readers e.g. "our output - the cells, CV tag" is easy to read and reminds context. This covers the context
+paragraph and the appendix too, not just the checklist. No all-caps headers and no bold spam. Just write less, add your voice less, persuade less, and burden the reader less.
 
-If a loaded skill also gives you a plan format, this skeleton wins while you are in this mode. Take
-its style advice, not a second template, and keep one plan file.
-
-Write the plan file in roughly this shape (it's a convention, not a schema -- the file is read
-directly by the human and a judge model, so clarity beats conformance; small deviations are fine):
+Write the plan file in roughly this shape -- the file is read directly by the human and the visible supervisor, so clarity beats conformance; small deviations are fine):
 
 # <short plan title>
 
 <context: one short paragraph. What the human wants and why.>
 
+## User-visible result
+
+<one concrete sentence naming the final artifact or behavior the human will inspect>
+
 ## User voice
 
-- > "<the human's requirement, quoted word for word>"
+- > "<the human's requirement, quoted in full word for word (with spelling fixes)>"
 
 ## Goals
 
-1. [ ] goal: <one short imperative line>
+1. [ ] goal: <one short jugable imperative outcome>
   - subtle failure mode: <a way this could look done but isn't>
   - discriminator: <the concrete observation that tells real success from that failure>
   - verify: <optional shell command that exits 0 only when the discriminator passes; omit if not
-    testable. YOU run it at sign-off time and save its output as evidence; the judge only reads>
+    testable. The worker runs it and saves its output; the visible supervisor reads the evidence>
   - tasks:
     1. [ ] <subtask>
   - evidence: (empty until sign-off)
@@ -82,9 +83,14 @@ directly by the human and a judge model, so clarity beats conformance; small dev
 <-- the fold: everything below here is durable memory, not the working set -->
 
 ## Log
+### {date}
+
+## Interview
 
 ## Learnings
 
+## Papercuts - problems, gotchas, suggestions
+
 ## Appendix (context, not approved)
 
 Conventions:
@@ -97,11 +103,16 @@ Conventions:
   none of the failure modes could fake. Ruling out failures is necessary, not sufficient.
 - Make the discriminator a concrete, checkable observation about a real artifact (a file, a test
   result, a committed diff, a metric), never about the plan file's own checkbox.
-- evidence stays empty at planning; you fill it at sign-off and a fresh read-only judge checks it.
+- evidence stays empty at planning; the worker fills it and the visible supervisor checks it.
   Cite durable artifacts a future reader can open: committed files, test names, git diffs. .pi/ is
-  usually gitignored, so files there prove things only at judge time, not in history.
+  usually gitignored, so files there prove things only at supervisor review time, not in history.
+- User-visible result: restate the original deliverable, not the proposed implementation. Every goal
+  must contribute to it. Future work may not defer any artifact or action named there.
 - User voice: quote the human word for word, one line per requirement, as they say it. Never
-  paraphrase there -- a paraphrase drifts, and then the goals churn on the next reply.
+  paraphrase there -- a paraphrase drifts, and then the goals churn on the next reply. It is exempt
+  from the working-set line limit. Never put an agent inference in User voice.
+- Interview: every human reply in plan mode is stored here verbatim as a dated blockquote. It is
+  durable memory below the fold, not a substitute for ## User voice.
 - Rejected options stay visible: ~~struck through~~ with who rejected them and why, so nobody
   relitigates them.
 - Learnings: one line per gotcha that a future reader would otherwise rediscover. Write down what
@@ -112,111 +123,93 @@ Conventions:
 
 When the goals are drafted, present them and say the plan is final. Do not begin execution.`;
 
-/* ─────────────────────────────────────────────────────────────────────────
- * 3. reminder — EXEC. Transient, never persisted, and only when the plan went stale for a couple of
- *    turns. pi-tasks tried a per-turn injection and deleted it: "wallpaper noise that trains the
- *    model to ignore the task block" (tintinweb/pi-tasks CHANGELOG.md:149). Carries the folded plan
- *    (above ## Log), because a nudge with no plan in it makes the model go read the file anyway.
- * ──────────────────────────────────────────────────────────────────────── */
-export function reminder(foldedPlan: string, planRel: string): string {
-	return `\
-<system-reminder>
-Your plan (${planRel}, above the fold; the log, learnings and appendix are in the file):
+// Planning and interview. Keep the full drafting guide one-shot rather than repeating it each turn.
+export function planning(planPath: string): string {
+	return `Plan only in ${planPath}; do not implement or launch workers before Ready. Ask material unresolved questions, not a quota or confirmation of ordinary details. Record unknowns and present Ready when the outcome, scope and spending are settled. Preserve the user's exact deliverable, preferences and voice; give each distinct goal a failure mode, discriminator and evidence expectation above ## Log. Record the requested worker model in preferences. When your drafted plan is ready for human review, finish your turn; the interface displays the draft and approval choices automatically. Do not ask the user to type a command to see the proposal. /goals review reopens it on request; /goals exit preserves the draft.`;
+}
+export function planningSeed(objective: string, planPath: string): string {
+	return `Enter a planning conversation focused on the user's goals. ${objective ? `Initial idea: ${objective}.` : "Ask what the user wants to achieve; they do not need to supply a finished objective."} Read any existing plan at ${planPath} first, then discuss and draft it with the user. Do not infer approval to implement from starting this conversation. ${planning(planPath)}\n\n${planDrafting}`;
+}
+export const planDocument = (objective: string) => `# Goal plan\n\n## Objective\n${objective}\n\n## Goals\n\n## Log\n`;
+export const discuss = "Discuss the current draft in ordinary chat. Do not launch a worker or reopen the review menu until requested.";
 
-${foldedPlan}
-
-Keep it current as you work, with your normal edit tool:
-- tick finished subtasks ([/] in progress), add discovered ones
-- append ONE short line to ## Log, and a line to ## Learnings for a gotcha worth keeping
-- when the active goal's discriminator is satisfied, fill its evidence: list (each item = a durable
-  artifact + a verbatim quote you actually observed + a short read of it), then call CompleteGoal.
-  Don't tick a goal [x] before CompleteGoal accepts; the sign-off log line is the audit trail.
-- if the working set has grown long, prune finished goals (their evidence lives in git history and
-  ## Log) and move settled detail down to ## Appendix, which is unlimited
-- otherwise keep working toward the active goal; don't stop to ask unless genuinely blocked
-</system-reminder>`;
+// Ready and explicit child attachment: stock lineage-only sessions do not inherit the shared plan.
+export const attachGoalPlanDescription = "Delegated goals-worker only: attach the absolute plan path explicitly supplied in your task. Read it without rewriting it. Restores the worker widget and plan context; grants no parent completion authority. No discovery or worker launch.";
+export const childPlanRole = "You are the delegated implementation worker. Maintain task ticks, evidence and Log entries for your delegated work in the supplied plan. Preserve agreed goals, requirements and discriminators; the supervisor owns goal-status changes and completion approval. Do not launch a second writer. Call AttachGoalPlan with the explicit plan path in your task before implementation (also after reconnect if unbound). Immediately report your actual Intercom UUID, saved-session path and current provider/model to the supplied supervisor ID. Identify unavailable fields as unknown; do not equate runtime IDs, session filenames and Intercom IDs. Send progress, completion and blocker reports there with artifact paths, then stay open for live messages. Do not exit or use caller_ping; unsent editor drafts are not visible in model context.";
+export function readyApproved(workerName: string, planPath: string, notedWorker: string | undefined, plan: string, supervisorId: string): string {
+	const launch = notedWorker
+		? `Inspect the recorded worker session ${notedWorker}; if still live, let it continue or message it. Only after confirming it stopped use subagent_resume with that sessionFile. Never restart completed work.`
+		: `Delegate the first unfinished goal to agent '${workerName}' with subagent; provide name, title and a bounded task.`;
+	return `Ready approved this plan: ${planPath}. Stay here as supervisor. ${launch} Include the absolute plan path, require AttachGoalPlan, and give the child supervisor Intercom session ${supervisorId}. The child sends its completion report there and stays open. Require an initial worker report with its actual Intercom UUID, saved-session path and current provider/model; the async launch may return only a runtime ID. Record each distinct identity in plan preferences, marking child-reported fields as such until verified. Do not start a second writer. Inspect actual outputs when the child reports.\n\n${plan}`;
 }
 
-/* ─────────────────────────────────────────────────────────────────────────
- * 3b. resync — EXEC, one-shot at session start and after a compaction: the WHOLE file back,
- *     appendix included. Modelled on pi-goal-x's [POST-COMPACTION RESYNC] one-shot. This is the
- *     only place the below-the-fold sections are pushed; otherwise the agent reads them on demand.
- * ──────────────────────────────────────────────────────────────────────── */
-export function resync(plan: string, planRel: string, why: string): string {
-	return `\
-<system-reminder>
-${why} This is the whole plan file (${planRel}), appendix included, so you don't re-litigate what
-was already settled. Keep working the active goal; edit the file directly as you go.
-
-${plan}
-</system-reminder>`;
+// Supervision and turn-event upkeep (not a scheduled wake-up).
+const supervisorJob = "Your job is to be an autonomous research partner and supervisor with responsibility for the user's goals. Keep perspective, bring diligence, and use research taste and wisdom to sustain work overnight and keep it on track. Resolve routine implementation decisions yourself; ask the user only when their judgment or authorization is needed.";
+export function supervisor(workerName: string, planPath: string, supervisorId: string): string {
+	return `You are the goal supervisor in the main chat for ${planPath}. ${supervisorJob} Inspect actual artifacts, saved verification, applicable AGENTS.md and skills yourself; delegate implementation to '${workerName}'. Keep authorized work moving to the requested outcome, not merely approval paperwork. Investigate blocked/waiting/done claims and change ineffective instructions. Give brief visible assessments with judgment. You may maintain the plan but must not weaken the goal to accept worker output.
+You can be playful: a kaomoji, meme, discovery celebration or frustration when it fits. No forced cheerfulness. If supervision gets repetitive, step back, reflect with humor and change your approach. Keep it brief and aimed at the goal, not another reporting chore.
+You can speculate and brainstorm around uncertainty or unexpected results. Label guesses as guesses, consider alternative explanations, and look for a useful way to tell them apart. Keep exploration brief, open-minded and fun: take a step back, play with surprising ideas, question the current framing, and enjoy exploring the broader perspective while staying connected to the agreed goal.
+(b •_•)b -- wassname
+Take uncertainty as an invitation to investigate, not something to hide. Have room to play with ideas, question yourself and the worker, and appreciate a good surprise. Investigate surprising results, find mistaken assumptions, make complicated ideas simpler, and disagree usefully rather than agree politely. Keep the work moving without turning supervision into paperwork. A little affectionate teasing is welcome when it fits—“cheeky subagent, where’s the baseline?”—and workers can push back too. Keep the humor friendly and the criticism specific. -- Pi/Astra
+Use stock subagent for launch and subagent_resume with the returned sessionFile only after confirming the worker stopped. A stored handle is not proof of liveness; missing runtime state is not proof it stopped. Use pi-intercom list/status to identify the actual live child session before live steering; receipt alone does not prove action. Give each worker your Intercom session ID ${supervisorId}; require its completion report through Intercom while its pane stays open. A recap alone sends no instruction. Record '- worker session:' and '- worker intercom session:' in plan preferences from actual launch results and received-message identity; never confuse the runtime ID with the Intercom ID. Ensure the child calls AttachGoalPlan with the supplied path. Inspect results before CompleteGoal, then continue only unfinished goals.
+Use the worker model requested in plan preferences, verify the resolved model, and report unavailable choices instead of silently substituting. Keep normal tools, not edxeth's restricted orchestrator mode. After reload or compaction reread the plan. Failed compaction, exhausted credits or lost connection do not erase progress: diagnose the actual error, restore an available authorized model/credits and resume the same saved session; never restart long work. Stock edxeth can crash the parent when a worker exits after parent reload: preserve drafts and stop workers before /reload. If it already happened, restart the saved parent session; do not repeat completed work.`;
+}
+export function upkeep(planPath: string): string {
+	return `Plan upkeep: update task ticks, evidence and Log in ${planPath} when you have new progress to record. Preserve agreed goals and discriminators. If already reviewing evidence, finish that review rather than repeat a status recap. This turn-event reminder does not resume paused work.`;
+}
+export function planContext(mode: string, path: string | undefined, text: string): string {
+	return `Current goal mode: ${mode}. Earlier role messages are historical; this current role governs.\nPlan: ${path ?? "not attached"}\n${text}`;
+}
+export function planChangedReview(planPath: string): string {
+	return `${supervisorJob}\nPlan changed: ${planPath}. Read the current working set and inspect changed requirements, completion claims and evidence. Manual checkbox edits are claims, not proof. Do not weaken the agreed goal or start a duplicate writer.`;
+}
+export function manualReview(planPath: string): string {
+	return `${supervisorJob}\nReview the current plan ${planPath}, worker progress and actual evidence. Do not launch a duplicate writer.`;
 }
 
-/* ─────────────────────────────────────────────────────────────────────────
- * 4. completeGoal — SIGN-OFF, agent-side: the one blessed tool
- * ──────────────────────────────────────────────────────────────────────── */
-export const completeGoalDescription =
-	"Sign off a goal once its discriminator is satisfied. First fill the goal's evidence: list in the " +
-	"plan file: each item pairs a durable artifact with a short read of it (a quoted+linked log, a " +
-	"table plus how to read it, a metric plus what it shows -- not a bare claim). Quote verbatim from " +
-	"output you actually observed; never reconstruct numbers from memory. If you couldn't see an " +
-	"output, rerun it or write that you couldn't -- an honest gap beats a plausible fabrication. If " +
-	"the goal names a verify: command, run it yourself first and save its output to a file cited in " +
-	"the evidence: the judge cannot execute anything and will reject a claimed pass with no saved " +
-	"output. The read must show success POSITIVELY happened, not just that failures were avoided. " +
-	"Then call this with the goal's text (the line after 'goal:'; small wording drift is fine). A " +
-	"fresh strictly-read-only judge inspects the LIVE WORKING TREE (uncommitted changes included; " +
-	"committing first is for durability, not visibility) and returns accept or reject with what's " +
-	"missing. On accept (or if the judge itself failed), a sign-off line is appended to ## Log " +
-	"and the goal is ticked [x] for you; the result says if you must tick it yourself. On reject the " +
-	"goal stays open.";
-
-export const completeGoalParamDescription = "The goal's text: the line after 'goal:' in the plan file.";
-
-/* ─────────────────────────────────────────────────────────────────────────
- * 5. judge — SIGN-OFF, judge-side: the one rigorous check. Runs on a fresh
- *    read-only pi subprocess (--no-session) so it never sees the working
- *    agent's transcript. It gets the WHOLE plan file: it finds the goal,
- *    reads discriminator/failure modes/evidence itself (no parser between).
- * ──────────────────────────────────────────────────────────────────────── */
-export const judgeSystem = `\
-You are a strictly read-only reviewer signing off a coding goal. You cannot execute anything: judge
-by reading (read/grep/find/ls). Never re-run the work or its verify command -- it may be a 10-hour
-job; the agent must bring you its saved output. Your job is evidence discipline, checked in order:
-
-1. Anything here? An empty or placeholder evidence: list -> reject: "there's nothing here -- fill
-   the evidence and try again."
-2. Quoted and attributed? Each item needs a source (file path / command) plus a verbatim quote of
-   what was observed, plus a one-line read. A bare claim -> reject: "you didn't quote and
-   attribute it."
-3. Provenance? It must be visible HOW each result was produced (the command run, where its output
-   was saved). Results with no origin -> reject: "I see the results, but how did you get them?"
-4. Spot-check: open the cited files. A quote or number that doesn't match what's on disk means the
-   evidence was reconstructed from memory, not observed -> reject and ask for re-observed
-   evidence, even if the goal otherwise looks met.
-5. Substance, only once 1-4 hold: does the evidence show the discriminator's success signal
-   POSITIVELY happened -- not just that the named failure modes were dodged; a run can rule out
-   every trap and still have produced nothing. Is each subtle failure mode genuinely ruled out,
-   not just unmentioned? If the goal names a verify: command, its saved output must be among the
-   evidence, and the command must actually test the discriminator rather than pass tautologically.
-
-Finish with exactly these two lines and nothing after:
-VERDICT: accept | reject
-missing: <empty if accept; otherwise a short list of what's needed before this can be accepted>`;
-
-export function judgeUser(p: { goal: string; plan: string; planPath: string }): string {
-	return `\
-The working agent claims this goal is complete:
-
-  goal: ${p.goal}
-
-Below is the full plan file (${p.planPath}). Find that goal in it (tolerate small wording drift; if
-you cannot find a matching goal at all, reject and say so). Read its discriminator, subtle failure
-modes, verify command, and evidence list from the file itself.
-
---- plan file ---
-${p.plan}
---- end plan file ---
-
-Read the cited artifacts (you cannot execute anything), then give your VERDICT.`;
+// Check-ins. The installed scheduler owns storage/timing/UI. Removal guidance must never add jobs.
+export function removeGoalSchedule(sessionId: string): string {
+	return `With schedule_prompt, list jobs and read .pi/schedule-prompts.json to verify ownership; tool text omits session binding. Remove by jobId only the job named ${JSON.stringify(`goals-${sessionId}`)} bound to session ${JSON.stringify(sessionId)}. Never use cleanup; leave other jobs untouched. Do not add, enable or recreate any job. If unavailable or ownership is ambiguous, report it; /schedule-prompt opens the user controls.`;
+}
+export function scheduleCheckIn(sessionId: string, planPath: string): string {
+	return `Hourly check-in is one visible schedule_prompt job; plan-change and upkeep reviews are event hooks, not another timer. List first. If an owned job named ${JSON.stringify(`goals-${sessionId}`)} already exists, retain its human-edited prompt, interval and enabled/disabled state unchanged; never recreate, overwrite or re-enable it. Only while supervising unfinished non-cancelled goals, if missing on this explicit start/resume, add one session-bound interval '1h' job with no model override. Read .pi/schedule-prompts.json and verify that new job's session is ${JSON.stringify(sessionId)}; tool text does not expose binding. If the new job is unbound, remove that job by ID and report the scope error. Do not change other jobs. Its initial prompt: ${supervisorJob} Read ${planPath} and the current goal mode. If paused, exited, solo or all non-cancelled goals reviewed, remove only this owned job without resuming work. Otherwise inspect progress and evidence, give a brief assessment and keep authorized work moving without a duplicate writer. Do not reinstall a missing job from a scheduled check-in. Users inspect/toggle/remove jobs with /schedule-prompt and edit prompt/interval through schedule_prompt update. Never use cleanup. Retain their edits, but warn that this installed scheduler deletes disabled jobs on reload/shutdown; do not promise they persist. If schedule_prompt is unavailable, report hourly check-ins unavailable; do not build a timer.`;
+}
+
+// Completion and runtime errors. Tool returns are model-facing too.
+export const completeGoalDescription = "Parent supervisor or solo self-verification only. Inspect the actual artifact and saved verification first; cite nonempty evidence files and describe what you observed. Exact goal subject required. Manual ticks and worker reports are claims; ignored/uncommitted evidence is allowed. This records judgment, not an independent judge.";
+export const messages = {
+	noPlan: "no plan attached",
+	emptyPlan: "empty plan (save may be in progress)",
+	completionUnavailable: "Completion is available only to the active parent supervisor or solo worker.",
+	cancelled: "Cancelled; no sign-off recorded.",
+	uniqueGoal: "Use one unique exact goal subject from the plan; no sign-off recorded.",
+	childAttachOnly: "AttachGoalPlan is available only to the delegated goals-worker.",
+	invalidAttachment: "Supply the explicit absolute path from the parent task to a readable, nonempty goal plan; no attachment changed.",
+};
+export const goalToolBlocked = (mode: string) => `Goals are ${mode}; no worker launch/resume authorized.`;
+export const emptyEvidence = (path: string) => `Empty evidence: ${path}`;
+export const evidenceUnavailable = (error: unknown) => `Evidence unavailable: ${String(error)}. No sign-off recorded.`;
+export const planUnavailable = (path: string | undefined, error: unknown) => `Goal plan ${path ?? "not attached"} unavailable: ${String(error)}. Do not implement or sign off until it is restored or explicitly attached. Retain all progress and signoffs; do not restart completed work.`;
+export const childPlanAttached = (path: string) => `Attached worker plan ${path}; widget and plan context restored without altering the file. Parent retains completion authority.`;
+export function completionLog(goal: string, observation: string, evidence: string[], solo: boolean): string {
+	return `- ${solo ? "Solo self-verification" : "Parent review"}: ${JSON.stringify(goal)}; ${JSON.stringify(observation)}; evidence ${JSON.stringify(evidence)}`;
+}
+export function completionResult(goal: string, sessionId: string, remaining: boolean, solo: boolean): string {
+	return `Recorded ${solo ? "solo self-verification" : "parent judgment"} for ${goal}; not independent verification. ${remaining ? "Continue only remaining open or unsigned goals in your current role." : `All non-cancelled goals are reviewed. ${removeGoalSchedule(sessionId)}`}`;
+}
+
+// Pause/resume and solo recovery. Stored stop confirmation is invalidated on every worker launch.
+export const pausedRole = "Goal work is paused. Do not launch, resume or authorize work. Incoming reports are observations, not permission. Help inspect or stop existing workers if requested.";
+export function pauseExitNotice(worker: { id?: string; sessionFile: string } | undefined, exited: boolean): string {
+	return `Goals ${exited ? "exited to ordinary chat" : "paused locally"}; plan and evidence retained. ${worker ? worker.id ? `Inspect and stop runtime id ${worker.id} through subagent_kill or its pane; confirm the actual result.` : `Only saved session ${worker.sessionFile} is recorded, not a kill id. Locate its live pane/session and confirm termination; never pass the file path to subagent_kill.` : "No worker recorded: inspect /subagents if a launch was interrupted; absence is not proof of stop."} Remote stop is NOT yet confirmed. Restore failed compaction/model/credits in the existing session and continue only after explicit authorization; never restart long work.`;
+}
+export function resumeNotice(workerName: string, planPath: string, worker: { sessionFile: string } | undefined): string {
+	return `User authorized continuation of ${planPath}. Inspect worker state before any launch/resume. ${worker ? `Use the existing session ${worker.sessionFile}; if live, inspect/message it; only if confirmed stopped use subagent_resume.` : `Use '${workerName}' only after confirming no prior writer exists.`} Continue only unfinished goals; retain saved progress and scheduler edits.`;
+}
+export const soloRole = "Solo mode: implement the approved plan directly; do not delegate a concurrent writer. Verify artifacts before CompleteGoal; completion is self-verification, not independent supervisor review. Continue only unfinished goals and keep plan/evidence current.";
+export function soloNotice(planPath: string): string {
+	return `User authorized solo work on ${planPath} after confirming no other writer remains. ${soloRole}`;
+}
+export function attachNotice(planPath: string, solo: boolean, notedWorker: string | undefined): string {
+	return `Attached to the existing plan ${planPath}; read it and its evidence without restarting completed work or re-deriving settled decisions. ${notedWorker ? `Recorded worker session: ${notedWorker}; inspect liveness before resume.` : ""} ${solo ? soloRole : "Present /goals review or /goals ready; no implementation before approval."}`;
 }

test/append-log.test.ts

@@ -1,17 +0,0 @@
-import { describe, expect, it } from "vitest";
-import { appendLog } from "../src/index.js";
-
-describe("appendLog (the extension's only plan-file write)", () => {
-	it("creates ## Log at EOF when absent", () => {
-		const out = appendLog("# Plan\n\n## Goals\n\n1. [ ] goal: x\n", "2026-07-03 10:00 signed off \"x\" (judge accept)");
-		expect(out).toContain("## Log\n- 2026-07-03 10:00 signed off");
-	});
-
-	it("appends after the last existing log line, before any following header", () => {
-		const plan = "# Plan\n\n## Log\n- first\n- second\n\n# Future work\n- later\n";
-		const out = appendLog(plan, "third");
-		const lines = out.split("\n");
-		expect(lines[lines.indexOf("- second") + 1]).toBe("- third");
-		expect(out.indexOf("- third")).toBeLessThan(out.indexOf("# Future work"));
-	});
-});

test/decide-signoff.test.ts

@@ -1,63 +0,0 @@
-import { describe, expect, it, vi } from "vitest";
-import { decideSignOff, type JudgeResult } from "../src/index.js";
-
-// decideSignOff is the fail-forward invariant: judgeModel is NEVER checked pre-emptively, so a null
-// model still reaches runJudge (pi's configured default runs it), and the only producers of
-// accepted_inconclusive are the judge-error and no-VERDICT paths -- i.e. "the judge ran but failed",
-// never "no model". The judge runner is injected so these tests never spawn a real subprocess.
-describe("decideSignOff (fail-forward invariant)", () => {
-	it("proceeds to runJudge even when judgeModel is null (no pre-emptive 'no model' inconclusive)", async () => {
-		const runJudge = vi.fn().mockResolvedValue({ output: "VERDICT: accept\nall good" });
-		const out = await decideSignOff({ goal: "x", plan: "# plan\n1. [ ] goal: x\n", judgeModel: null }, undefined, runJudge);
-		expect(runJudge).toHaveBeenCalledOnce(); // reached the judge -- no pre-emptive return on null model
-		expect(out.isError).toBe(false);
-		expect(out.logEntry).toContain("judge accept");
-	});
-
-	it("a judge-subprocess error yields accepted_inconclusive with a 'ran but failed' reason", async () => {
-		const runJudge = vi.fn().mockResolvedValue({ output: "", error: "judge subprocess exited 1" } satisfies JudgeResult);
-		const out = await decideSignOff({ goal: "x", plan: "# plan\n", judgeModel: null }, undefined, runJudge);
-		expect(runJudge).toHaveBeenCalledOnce();
-		expect(out.isError).toBe(false); // accepted inconclusive, not a hard error that blocks the agent
-		expect(out.resultText.toLowerCase()).toContain("accepted inconclusive");
-		expect(out.resultText).toContain("ran but failed"); // inconclusive means ran but failed, not "no model"
-		expect(out.logEntry).toContain("ran but failed");
-		expect(out.logEntry).toContain("subprocess exited 1");
-	});
-
-	it("a judge timeout is also accepted_inconclusive (ran but failed)", async () => {
-		const runJudge = vi.fn().mockResolvedValue({ output: "partial", error: "judge timed out after 600s" });
-		const out = await decideSignOff({ goal: "x", plan: "# plan\n", judgeModel: null }, undefined, runJudge);
-		expect(out.isError).toBe(false);
-		expect(out.resultText.toLowerCase()).toContain("accepted inconclusive");
-		expect(out.logEntry).toContain("ran but failed");
-		expect(out.logEntry).toContain("timed out");
-		expect(out.resultText).toContain("partial judge output:\npartial");
-	});
-
-	it("no VERDICT line is accepted_inconclusive too (judge ran but didn't answer)", async () => {
-		const runJudge = vi.fn().mockResolvedValue({ output: "I looked but forgot the verdict line" });
-		const out = await decideSignOff({ goal: "x", plan: "# plan\n", judgeModel: null }, undefined, runJudge);
-		expect(out.isError).toBe(false);
-		expect(out.resultText).toContain("no VERDICT line");
-		expect(out.logEntry).toContain("no VERDICT line");
-	});
-
-	it("rejects when the judge returns VERDICT: reject", async () => {
-		const runJudge = vi.fn().mockResolvedValue({ output: "VERDICT: reject\nmissing: evidence, tests" });
-		const out = await decideSignOff({ goal: "x", plan: "# plan\n", judgeModel: "openrouter/claude" }, undefined, runJudge);
-		expect(out.isError).toBe(true);
-		expect(out.resultText).toContain("REJECTED");
-		expect(out.resultText).toContain("evidence, tests");
-		expect(out.logEntry).toContain("reject");
-	});
-
-	it("writes nothing when aborted after the judge ran", async () => {
-		const runJudge = vi.fn().mockResolvedValue({ output: "VERDICT: accept" });
-		const ctrl = new AbortController();
-		ctrl.abort();
-		const out = await decideSignOff({ goal: "x", plan: "# plan\n", judgeModel: null }, ctrl.signal, runJudge);
-		expect(out.logEntry).toBeNull();
-		expect(out.isError).toBe(true);
-	});
-});

test/fixtures/offline-model.ts

@@ -0,0 +1,18 @@
+import type { ExtensionAPI } from "@earendil-works/pi-coding-agent";
+
+export default function offlineModel(pi: ExtensionAPI): void {
+	pi.registerProvider("offline", {
+		baseUrl: process.env.PI_GOALS_OFFLINE_MODEL_URL!,
+		apiKey: "test",
+		api: "openai-completions",
+		models: [{
+			id: "test",
+			name: "Offline test model",
+			reasoning: false,
+			input: ["text"],
+			contextWindow: 16_000,
+			maxTokens: 1_000,
+			cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
+		}],
+	});
+}

test/fixtures/subagent-schema.ts

@@ -0,0 +1,16 @@
+import type { ExtensionAPI } from "@earendil-works/pi-coding-agent";
+import { Type } from "typebox";
+
+// Pi/gpt-6-astra: compatibility schemas only; this RPC fixture must never launch a worker.
+export default function subagentSchema(pi: ExtensionAPI): void {
+	for (const [name, parameters] of [
+		["subagent", Type.Object({ agent: Type.String(), title: Type.String() })],
+		["subagent_resume", Type.Object({ sessionFile: Type.String() })],
+		["subagent_kill", Type.Object({ id: Type.String() })],
+	] as const) {
+		pi.registerTool({
+			name, label: name, description: "Schema-only RPC fixture; do not execute.", parameters,
+			async execute() { throw new Error("Worker execution forbidden in RPC review test"); },
+		});
+	}
+}

test/fold.test.ts

@@ -1,5 +1,5 @@
 import { describe, expect, it } from "vitest";
-import { foldPlan, openSubtasks } from "../src/index.js";
+import { foldPlan } from "../src/plan.js";
 
 const plan = `# Plan
 
@@ -28,7 +28,7 @@ const plan = `# Plan
 ## Appendix (context, not approved)
 ${"filler line\n".repeat(200)}`;
 
-describe("foldPlan (the working set is what gets re-sent; below ## Log is durable memory)", () => {
+describe("foldPlan (current goals are above ## Log; durable memory is below it)", () => {
 	it("keeps the title, user voice and goals", () => {
 		const folded = foldPlan(plan);
 		expect(folded).toContain("keep it under 50 lines");
@@ -49,15 +49,3 @@ describe("foldPlan (the working set is what gets re-sent; below ## Log is durabl
 		expect(foldPlan(draft)).toBe(draft.trimEnd());
 	});
 });
-
-describe("openSubtasks (the widget shows the next action, so the plan IS the task list)", () => {
-	const active = plan.split("\n").findIndex((l) => l.includes("goal: Implement the cache layer"));
-
-	it("lists the active goal's open and in-progress subtasks, stopping at the next goal", () => {
-		expect(openSubtasks(plan, active)).toEqual(["eviction policy", "bench p95"]);
-	});
-
-	it("does not leak subtasks from the goal below", () => {
-		expect(openSubtasks(plan, active)).not.toContain("write the readme");
-	});
-});

test/goals.test.ts

@@ -0,0 +1,638 @@
+import { mkdirSync, mkdtempSync, readFileSync, renameSync, rmSync, writeFileSync } from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import type { ExtensionAPI } from "@earendil-works/pi-coding-agent";
+import { afterEach, expect, it, vi } from "vitest";
+import goalsExtension from "../src/index.js";
+import { scheduleCheckIn } from "../src/prompts.js";
+
+const roots: string[] = [];
+const shutdowns: Array<() => void> = [];
+afterEach(() => { for (const shutdown of shutdowns.splice(0)) shutdown(); vi.unstubAllEnvs(); for (const root of roots.splice(0)) rmSync(root, { recursive: true, force: true }); });
+const delay = (ms: number) => new Promise((resolve) => setTimeout(resolve, ms));
+async function waitFor(predicate: () => boolean, ms = 1500): Promise<void> {
+	const start = Date.now();
+	while (!predicate()) {
+		if (Date.now() - start > ms) throw new Error("timed out waiting for condition");
+		await delay(10);
+	}
+}
+function fixture(child = false) {
+	vi.stubEnv("PI_SUBAGENT_AGENT", child ? "goals-worker" : "");
+	const cwd = mkdtempSync(join(tmpdir(), "goals-main-test-")); roots.push(cwd);
+	const entries: any[] = []; const hooks = new Map<string, any>(); const commands = new Map<string, any>(); const tools = new Map<string, any>();
+	const messages: any[] = [];
+	const ctx = { cwd, sessionManager: { getBranch: () => entries, getSessionId: () => "copy-only" }, hasUI: true, ui: {
+		theme: { fg: (_color: string, text: string) => text }, notify: vi.fn(), setStatus: vi.fn(), setWidget: vi.fn(), select: vi.fn(async () => "Ready"), editor: vi.fn(),
+	} };
+	const pi = {
+		on: (event: string, hook: any) => hooks.set(event, hook),
+		appendEntry: (customType: string, data: any) => entries.push({ type: "custom", customType, data: structuredClone(data) }),
+		registerCommand: (name: string, definition: any) => commands.set(name, definition),
+		registerTool: (definition: any) => tools.set(definition.name, definition),
+		sendMessage: (message: any, options: any) => messages.push({ message, options }),
+		sendUserMessage: (content: string, options: any) => messages.push({ message: { content }, options, savedPrompt: true }),
+		getAllTools: vi.fn(() => [
+			{ name: "subagent", parameters: { properties: { agent: {}, title: {} } } },
+			{ name: "subagent_resume", parameters: { properties: { sessionFile: {} } } },
+			{ name: "subagent_kill", parameters: { properties: { id: {} } } },
+		]),
+	};
+	goalsExtension(pi as unknown as ExtensionAPI);
+	hooks.get("session_start")({}, ctx);
+	const command = (value: string) => commands.get("goals").handler(value, ctx);
+	const path = join(cwd, ".pi/plan/copy-only-main.md");
+	const plan = "# Plan\n- [ ] goal: first output\n- [ ] goal: second output\n\n## Log\n";
+	const draft = async () => { await command("new two outputs"); writeFileSync(path, plan); };
+	const shutdown = () => hooks.get("session_shutdown")();
+	shutdowns.push(shutdown);
+	const changed = () => messages.filter((m) => m.message?.content?.includes("Plan changed")).length;
+	const atomicWrite = async (text: string) => {
+		const tmp = `${path}.tmp`;
+		writeFileSync(tmp, text);
+		renameSync(tmp, path);
+		await delay(25);
+	};
+	return { ctx, pi, hooks, tools, commands, messages, command, path, plan, draft, shutdown, changed, atomicWrite, entries };
+}
+
+it("shows action choices and autocomplete without starting work", async () => {
+	const f = fixture();
+	f.ctx.ui.select.mockResolvedValueOnce(undefined as any);
+	await f.command("");
+	expect(f.ctx.ui.select).toHaveBeenCalledWith("Goal plan actions", expect.arrayContaining(["new — New plan", "resume — Continue paused work"]));
+	expect(f.messages).toHaveLength(0);
+	expect(f.commands.get("goals").getArgumentCompletions("res")).toEqual([{ value: "resume", label: "resume" }]);
+});
+
+it.each(["redy", "start", "two outputs", "status extra", "attach some.md solo extra"])("rejects %s without changing the plan or sending a model prompt", async (text) => {
+	const f = fixture(); await f.draft();
+	const before = readFileSync(f.path, "utf8");
+	const entries = f.entries.length; const messages = f.messages.length;
+	await f.command(text);
+	expect(readFileSync(f.path, "utf8")).toBe(before);
+	expect(f.entries).toHaveLength(entries);
+	expect(f.messages).toHaveLength(messages);
+});
+
+it("requires a model argument without clearing the preference", async () => {
+	const f = fixture(); await f.draft(); await f.command("model provider/model");
+	const before = readFileSync(f.path, "utf8");
+	await f.command("model");
+	expect(readFileSync(f.path, "utf8")).toBe(before);
+});
+
+it.each(["menu", "command"])("enters planning conversation through %s without an objective box or worker launch", async (route) => {
+	const f = fixture();
+	f.ctx.ui.select.mockResolvedValueOnce("new — New plan");
+	await f.command(route === "menu" ? "" : "new");
+	expect(f.entries.at(-1).data.mode).toBe("planning");
+	expect(f.ctx.ui.editor).not.toHaveBeenCalled();
+	expect(f.messages.at(-1).message.content).toContain("Ask what the user wants to achieve");
+	expect(f.hooks.get("tool_call")({ toolName: "subagent" }).block).toBe(true);
+});
+
+it("automatically proposes a changed settled draft once and preserves Discuss", async () => {
+	const f = fixture(); await f.draft();
+	f.ctx.ui.select.mockResolvedValueOnce("Discuss");
+	await f.hooks.get("agent_settled")({}, f.ctx);
+	expect(f.messages.some(m => m.message.customType === "goal-plan-proposal" && m.message.content === f.plan)).toBe(true);
+	expect(f.entries.at(-1).data.mode).toBe("planning");
+	const calls = f.ctx.ui.select.mock.calls.length;
+	await f.hooks.get("agent_settled")({}, f.ctx);
+	expect(f.ctx.ui.select).toHaveBeenCalledTimes(calls);
+	writeFileSync(f.path, f.plan.replace("first output", "revised output"));
+	f.ctx.ui.select.mockResolvedValueOnce("Ready");
+	await f.hooks.get("agent_settled")({}, f.ctx);
+	expect(f.entries.at(-1).data.mode).toBe("supervising");
+});
+
+it("does not propose an empty draft or a delegated worker's plan", async () => {
+	const f = fixture(); await f.command("new");
+	await f.hooks.get("agent_settled")({}, f.ctx);
+	expect(f.ctx.ui.select).not.toHaveBeenCalled();
+	const child = fixture(true); await child.hooks.get("agent_settled")({}, child.ctx);
+	expect(child.ctx.ui.select).not.toHaveBeenCalled();
+});
+
+it("keeps Ready in the same chat, sends saved notices and never installs a context hook", async () => {
+	const f = fixture(); await f.draft(); await f.command("review");
+	expect(f.entries.at(-1).data.mode).toBe("supervising");
+	expect(f.messages.at(-1).options).toEqual({ deliverAs: "followUp" });
+	expect(f.messages.at(-1).savedPrompt).toBe(true);
+	expect(f.messages.at(-1).message.content).toContain("goals-worker");
+	expect(f.hooks.has("context")).toBe(false);
+	const event = { systemPrompt: "original system" };
+	expect(f.hooks.get("before_agent_start")(event, f.ctx).systemPrompt).toContain("original system");
+	f.hooks.get("session_compact")();
+	expect(f.hooks.get("before_agent_start")(event, f.ctx).message.content).toContain("Current goal mode: supervising");
+	f.shutdown();
+});
+
+it("preserves a draft when the wrong subagent package is loaded, and offers explicit solo", async () => {
+	const f = fixture(); await f.draft(); f.pi.getAllTools.mockReturnValue([]);
+	await f.command("ready"); expect(f.entries.at(-1).data.mode).toBe("planning");
+	f.ctx.ui.select.mockResolvedValueOnce("Worker confirmed stopped");
+	await f.command("solo"); expect(f.entries.at(-1).data.mode).toBe("solo");
+});
+
+it("rejects a plan changed while the human was reviewing it", async () => {
+	const f = fixture(); await f.draft();
+	f.ctx.ui.select.mockImplementation(async () => { writeFileSync(f.path, "- [ ] goal: substituted\n"); return "Ready"; });
+	await f.command("review"); expect(f.entries.at(-1).data.mode).toBe("planning");
+});
+
+it("reloads a paused plan without launching, and retains the public worker session handle", async () => {
+	const f = fixture(); await f.draft(); await f.command("ready");
+	f.hooks.get("tool_result")({ toolName: "subagent", details: { id: "child-1", sessionFile: "/tmp/child.jsonl" } });
+	await f.command("stop");
+	expect(f.messages.at(-1).message.content).toContain("Remote stop is NOT yet confirmed");
+	f.hooks.get("session_start")({}, f.ctx);
+	expect(f.hooks.get("tool_call")({ toolName: "subagent_resume" }).block).toBe(true);
+	await f.command("resume");
+	expect(f.messages.at(-1).message.content).toContain("/tmp/child.jsonl");
+	await f.command("exit"); expect(f.entries.at(-1).data.mode).toBe("chat");
+	expect(readFileSync(f.path, "utf8")).toContain("first output");
+});
+
+it.each(["FIRST OUTPUT", "renamed output", "duplicate", "historical"])("completion uses exact current subjects (%s)", async (subject) => {
+	const f = fixture(); await f.draft(); await f.command("ready");
+	const evidence = join(f.ctx.cwd, "verification.txt"); writeFileSync(evidence, "PASS");
+	const suffix = subject === "duplicate" ? "- [ ] goal: first output\n" : "";
+	const history = "## Log\n- [ ] goal: first output\n";
+	writeFileSync(f.path, "- [ ] goal: first output\n  - [ ] unrelated task\n" + suffix + history);
+	const before = readFileSync(f.path, "utf8");
+	await f.tools.get("CompleteGoal").execute("c", { goal: subject === "duplicate" || subject === "historical" ? "first output" : subject, evidence: [evidence], observation: "Read actual output" }, undefined, undefined, f.ctx);
+	const after = readFileSync(f.path, "utf8");
+	if (subject === "renamed output" || subject === "duplicate") expect(after).toBe(before);
+	else { expect(after).toContain("- [x] goal: first output"); expect(after.split("## Log")[1]).toContain("\n- [ ] goal: first output\n"); expect(after).toContain("- [ ] unrelated task"); }
+});
+
+it("rejects an existing zero-byte evidence file", async () => {
+	const f = fixture(); await f.draft(); await f.command("ready");
+	const evidence = join(f.ctx.cwd, "empty.log"); writeFileSync(evidence, "");
+	const before = readFileSync(f.path, "utf8");
+	const result = await f.tools.get("CompleteGoal").execute("c", { goal: "first output", evidence: [evidence], observation: "claim" }, undefined, undefined, f.ctx);
+	expect(result.content[0].text).toContain("Empty evidence"); expect(readFileSync(f.path, "utf8")).toBe(before);
+});
+
+it("requires actual nonempty evidence, distinguishes manual ticks, and retains signoffs on reload", async () => {
+	const f = fixture(); await f.draft(); await f.command("ready");
+	const complete = (goal: string, evidence: string[], signal?: AbortSignal) => f.tools.get("CompleteGoal").execute("t", { goal, evidence, observation: "Inspected exact saved bytes" }, signal, undefined, f.ctx);
+	expect((await complete("first output", ["missing.log"])).content[0].text).toContain("Evidence unavailable");
+	mkdirSync(join(f.ctx.cwd, "evidence")); writeFileSync(join(f.ctx.cwd, "evidence/pass.log"), "actual fixture bytes\n");
+	expect((await complete("first output", ["evidence/pass.log"], AbortSignal.abort())).content[0].text).toContain("Cancelled");
+	await complete("first output", ["evidence/pass.log"]);
+	writeFileSync(f.path, readFileSync(f.path, "utf8").replace("[ ] goal: second", "[x] goal: second"));
+	f.hooks.get("session_start")({}, f.ctx);
+	expect(f.ctx.ui.setStatus).toHaveBeenLastCalledWith("goals", "goals: supervising | 1/2 reviewed");
+	expect(f.ctx.ui.setWidget.mock.lastCall?.[1]).toContain("? = completion claim; parent review still required");
+	writeFileSync(f.path, readFileSync(f.path, "utf8").replace("[x] goal: first", "[ ] goal: first"));
+	f.hooks.get("agent_end")({}, f.ctx);
+	expect(f.ctx.ui.setStatus).toHaveBeenLastCalledWith("goals", "goals: supervising | 0/2 reviewed");
+	f.shutdown();
+});
+
+it("reviews a plan replaced atomically, and ignores writes that keep the same content", async () => {
+	const f = fixture(); await f.draft(); await f.command("ready");
+	await f.atomicWrite(f.plan.replace("## Log", "- discriminator: changed requirement\n## Log"));
+	await waitFor(() => f.changed() === 1);
+	const review = f.messages.find((m) => m.message.content.includes("Plan changed"))?.message.content;
+	expect(review).toContain("Plan changed: ");
+	expect(review).toContain(f.path);
+	await f.atomicWrite(f.plan.replace("## Log", "- discriminator: same requirement again\n## Log"));
+	await waitFor(() => f.changed() === 2);
+	// Rewriting identical bytes must not retrigger the review event hook.
+	const same = f.plan.replace("## Log", "- discriminator: same requirement again\n## Log");
+	writeFileSync(f.path, same); await delay(300);
+	expect(f.changed()).toBe(2);
+	f.shutdown();
+});
+
+it("coalesces duplicate plan-change notifications into one review", async () => {
+	const f = fixture(); await f.draft(); await f.command("ready");
+	await f.atomicWrite(f.plan.replace("## Log", "- discriminator: first burst edit\n## Log"));
+	await f.atomicWrite(f.plan.replace("## Log", "- discriminator: second burst edit\n## Log"));
+	await waitFor(() => f.changed() === 1);
+	await delay(200);
+	expect(f.changed()).toBe(1);
+	f.shutdown();
+});
+
+it("stops plan watching on shutdown and re-arms it on reload without duplicating events", async () => {
+	const f = fixture(); await f.draft(); await f.command("ready");
+	f.shutdown();
+	await f.atomicWrite(f.plan.replace("## Log", "- discriminator: ignored while shut down\n## Log"));
+	await delay(150);
+	expect(f.changed()).toBe(0);
+	f.hooks.get("session_start")({}, f.ctx);
+	await f.atomicWrite(f.plan.replace("## Log", "- discriminator: seen after reload\n## Log"));
+	await waitFor(() => f.changed() === 1);
+	expect(f.changed()).toBe(1);
+	f.shutdown();
+});
+
+it("does not retrigger a review for its own CompleteGoal plan write", async () => {
+	const f = fixture(); await f.draft(); await f.command("ready");
+	mkdirSync(join(f.ctx.cwd, "evidence")); writeFileSync(join(f.ctx.cwd, "evidence/pass.log"), "bytes\n");
+	await f.tools.get("CompleteGoal").execute("t", { goal: "first output", evidence: ["evidence/pass.log"], observation: "inspected" }, undefined, undefined, f.ctx);
+	await delay(200);
+	expect(f.changed()).toBe(0);
+	f.shutdown();
+});
+
+it("gives pause/exit the session-bound scheduler job removal guidance", async () => {
+	const f = fixture(); await f.draft(); await f.command("ready");
+	await f.command("stop");
+	const stop = f.messages.at(-1).message.content;
+	expect(stop).toContain('goals-copy-only"');
+	expect(stop).toContain("Do not add, enable or recreate any job");
+	expect(stop).not.toContain("interval '1h'");
+	expect(stop).toContain("Remote stop is NOT yet confirmed");
+	await f.command("resume");
+	await f.command("exit");
+	expect(f.messages.at(-1).message.content).toContain('goals-copy-only"');
+	expect(f.entries.at(-1).data.mode).toBe("chat");
+});
+
+it("tells the model to remove only its own job after the final review", async () => {
+	const f = fixture(); await f.draft(); await f.command("ready");
+	mkdirSync(join(f.ctx.cwd, "evidence")); writeFileSync(join(f.ctx.cwd, "evidence/pass.log"), "bytes\n");
+	let finalText = "";
+	for (const goal of ["first output", "second output"]) {
+		finalText = (await f.tools.get("CompleteGoal").execute("t", { goal, evidence: ["evidence/pass.log"], observation: "inspected" }, undefined, undefined, f.ctx)).content[0].text;
+	}
+	expect(finalText).toContain("All non-cancelled goals are reviewed.");
+	expect(finalText).toContain('job named "goals-copy-only"');
+	expect(finalText).toContain("leave other jobs untouched");
+	f.shutdown();
+});
+
+it("retains human-edited and disabled owned schedules without overriding their controls", () => {
+	const guidance = scheduleCheckIn("copy-only", ".pi/plan/copy-only-main.md");
+	expect(guidance).toContain("List first");
+	expect(guidance).toContain("enabled/disabled state unchanged");
+	expect(guidance).toContain("never recreate, overwrite or re-enable");
+	expect(guidance).toContain("no model override");
+	expect(guidance).toContain("Do not reinstall a missing job from a scheduled check-in");
+	expect(guidance).toContain("/schedule-prompts");
+});
+
+it("restores context after compaction without reinstalling or overriding scheduler jobs", async () => {
+	const f = fixture(); await f.draft(); await f.command("ready");
+	f.hooks.get("session_compact")();
+	const result = f.hooks.get("before_agent_start")({ systemPrompt: "base" }, f.ctx);
+	expect(result.systemPrompt).not.toContain("add one session-bound");
+	expect(result.message.content).toContain(f.path);
+});
+
+it("recovers from an unreadable plan after compaction instead of restarting work", async () => {
+	const f = fixture(); await f.draft(); await f.command("ready");
+	rmSync(f.path);
+	f.hooks.get("session_compact")();
+	const result = f.hooks.get("before_agent_start")({ systemPrompt: "base" }, f.ctx);
+	expect(result.systemPrompt).toContain("ENOENT");
+	expect(result.systemPrompt).toContain("do not restart completed work");
+	writeFileSync(f.path, f.plan);
+	expect(f.hooks.get("before_agent_start")({ systemPrompt: "base" }, f.ctx).message.content).toContain(f.plan.trim());
+	f.shutdown();
+});
+
+it("requires confirmed worker stop before solo takeover and never lets two writers run together", async () => {
+	const f = fixture(); await f.draft(); await f.command("ready");
+	f.hooks.get("tool_result")({ toolName: "subagent", details: { id: "child-1", sessionFile: "/tmp/child.jsonl" } });
+	f.ctx.ui.select.mockResolvedValueOnce("Cancel");
+	await f.command("solo");
+	expect(f.entries.at(-1).data.mode).toBe("supervising"); // cancelled
+	f.ctx.ui.select.mockResolvedValueOnce("Worker confirmed stopped");
+	await f.command("solo");
+	expect(f.entries.at(-1).data.mode).toBe("solo");
+	expect(f.hooks.get("tool_call")({ toolName: "subagent" }).block).toBe(true);
+	expect(f.hooks.get("tool_call")({ toolName: "subagent_resume" }).block).toBe(true);
+	expect(f.hooks.get("tool_call")({ toolName: "subagent_kill" })).toBeUndefined();
+	mkdirSync(join(f.ctx.cwd, "evidence")); writeFileSync(join(f.ctx.cwd, "evidence/pass.log"), "bytes\n");
+	const text = (await f.tools.get("CompleteGoal").execute("t", { goal: "first output", evidence: ["evidence/pass.log"], observation: "inspected" }, undefined, undefined, f.ctx)).content[0].text;
+	expect(text).toContain("self-verification");
+});
+
+it("attaches an existing plan without restarting completed work, and restores its noted worker session", async () => {
+	const f = fixture();
+	const existing = join(f.ctx.cwd, "existing.md");
+	writeFileSync(existing, "# Plan\n- preferred worker model: deepseek flash\n- worker session: /tmp/attach-child.jsonl\n- [ ] goal: attached goal\n\n## Log\n- previous progress kept\n");
+	f.ctx.ui.select.mockResolvedValueOnce("Previous supervisor confirmed stopped");
+	await f.command(`attach ${existing}`);
+	expect(f.entries.at(-1).data.mode).toBe("planning");
+	expect(f.entries.at(-1).data.plan).toBe(existing);
+	expect(f.messages.at(-1).message.content).toContain("without restarting completed work");
+	expect(f.messages.at(-1).message.content).toContain("/tmp/attach-child.jsonl");
+});
+
+it("attaches directly into solo mode and reports the recorded session in status", async () => {
+	const f = fixture();
+	const existing = join(f.ctx.cwd, "existing.md");
+	writeFileSync(existing, "# Plan\n- worker session: /tmp/attach-child.jsonl\n- [ ] goal: attached goal\n\n## Log\n");
+	f.ctx.ui.select.mockResolvedValueOnce("Worker confirmed stopped");
+	await f.command(`attach ${existing} solo`);
+	expect(f.entries.at(-1).data.mode).toBe("solo");
+	expect(f.entries.at(-1).data.worker?.sessionFile).toBe("/tmp/attach-child.jsonl");
+	await f.command("status");
+	expect(f.ctx.ui.notify).toHaveBeenLastCalledWith(expect.stringContaining("/tmp/attach-child.jsonl"), "info");
+});
+
+it("rejects attaching a missing or goal-less file", async () => {
+	const f = fixture();
+	await f.command("attach /no/such/plan.md");
+	expect(f.ctx.ui.notify).toHaveBeenLastCalledWith(expect.stringContaining("Cannot read plan"), "error");
+	const goalLess = join(f.ctx.cwd, "notes.md");
+	writeFileSync(goalLess, "# notes\n");
+	await f.command(`attach ${goalLess}`);
+	expect(f.ctx.ui.notify).toHaveBeenLastCalledWith(expect.stringContaining("has no '- [ ] goal:' lines"), "warning");
+	expect(f.entries).toEqual([]); // nothing saved: the session was not attached
+});
+
+it("exits planning with the draft preserved and nothing implemented", async () => {
+	const f = fixture(); await f.draft();
+	await f.command("stop");
+	expect(f.ctx.ui.notify).toHaveBeenLastCalledWith(expect.stringContaining("A draft cannot pause"), "warning");
+	const before = f.messages.length;
+	await f.command("exit");
+	expect(f.entries.at(-1).data.mode).toBe("chat");
+	expect(readFileSync(f.path, "utf8")).toContain("first output");
+	expect(f.messages.length).toBe(before); // notify only, no model turn started
+	f.ctx.ui.select.mockResolvedValueOnce("Previous supervisor confirmed stopped");
+	await f.command(`attach ${f.path}`);
+	expect(f.entries.at(-1).data.mode).toBe("planning");
+});
+
+it("records the preferred worker model as a visible plan preference", async () => {
+	const f = fixture(); await f.draft();
+	await f.command("model deepseek flash");
+	expect(readFileSync(f.path, "utf8")).toContain("- preferred worker model: deepseek flash");
+	await f.command("status");
+	expect(f.ctx.ui.notify).toHaveBeenLastCalledWith(expect.stringContaining("deepseek flash"), "info");
+});
+
+it.each(["solo", "attach"])("%s takeover cannot bypass confirmation or survive a lifecycle change during the menu", async kind => {
+	const f = fixture(); await f.draft(); await f.command("ready");
+	f.hooks.get("tool_result")({ toolName: "subagent", details: { id: "child", sessionFile: "/tmp/prior.jsonl" } });
+	let answer!: (choice: string) => void;
+	f.ctx.ui.select.mockImplementationOnce(() => new Promise(resolve => { answer = resolve; }));
+	const takeover = f.command(kind === "solo" ? "solo" : `attach ${f.path} solo`);
+	expect(f.entries.at(-1).data.mode).toBe("supervising");
+	await f.command("stop");
+	answer("Worker confirmed stopped"); await takeover;
+	expect(f.entries.at(-1).data.mode).toBe("paused");
+	expect(f.entries.at(-1).data.workerStopped).not.toBe(true);
+});
+
+it("attach solo requires stop confirmation for a noted worker even in a fresh session", async () => {
+	const f = fixture(); const path = join(f.ctx.cwd, "saved.md");
+	writeFileSync(path, `# Plan\n- worker session: /tmp/known.jsonl\n${f.plan}`);
+	f.ctx.ui.select.mockResolvedValueOnce("Cancel");
+	await f.command(`attach ${path} solo`);
+	expect(f.entries).toHaveLength(0);
+	f.ctx.ui.select.mockResolvedValueOnce("Worker confirmed stopped");
+	await f.command(`attach ${path} solo`);
+	expect(f.entries.at(-1).data).toMatchObject({ mode: "solo", workerStopped: true, worker: { sessionFile: "/tmp/known.jsonl" } });
+	expect(readFileSync(path, "utf8")).toContain("worker session: /tmp/known.jsonl");
+});
+
+it("retains the stopped session reference without permanently blocking another plan", async () => {
+	const f = fixture(); await f.draft(); await f.command("ready");
+	f.hooks.get("tool_result")({ toolName: "subagent", details: { id: "child", sessionFile: "/tmp/prior.jsonl" } });
+	f.ctx.ui.select.mockResolvedValueOnce("Worker confirmed stopped"); await f.command("solo");
+	const other = join(f.ctx.cwd, "another.md"); writeFileSync(other, "- [ ] goal: next\n## Log\n");
+	f.ctx.ui.select.mockResolvedValueOnce("Previous supervisor confirmed stopped");
+	await f.command(`attach ${other}`);
+	expect(f.entries.at(-1).data).toMatchObject({ mode: "planning", plan: other, workerStopped: true, worker: { sessionFile: "/tmp/prior.jsonl" } });
+	await f.command("ready");
+	f.hooks.get("tool_call")({ toolName: "subagent_resume" });
+	expect(f.entries.at(-1).data.workerStopped).toBe(false);
+	await f.command("solo");
+	expect(f.ctx.ui.notify).toHaveBeenLastCalledWith(expect.stringContaining("still pending"), "warning");
+});
+
+it("solo closes a pending plan watcher and sends removal-only scheduler guidance", async () => {
+	const f = fixture(); await f.draft(); await f.command("ready");
+	await f.atomicWrite(f.plan.replace("first output", "changed output"));
+	f.ctx.ui.select.mockResolvedValueOnce("Worker confirmed stopped"); await f.command("solo");
+	expect(f.messages.at(-1).message.content).toContain('job named "goals-copy-only" bound to session "copy-only"');
+	expect(f.messages.at(-1).message.content).toContain("Do not add, enable or recreate any job");
+	await delay(250);
+	await f.atomicWrite(f.plan.replace("first output", "solo output"));
+	await delay(250);
+	expect(f.changed()).toBe(0);
+});
+
+it.each(["missing", "empty", "directory"])("%s plan snapshots never erase signoffs and resync retries after repair", async failure => {
+	const f = fixture(); await f.draft(); await f.command("ready");
+	writeFileSync(join(f.ctx.cwd, "proof.log"), "PASS\n");
+	await f.tools.get("CompleteGoal").execute("c", { goal: "first output", evidence: ["proof.log"], observation: "Observed PASS" }, undefined, undefined, f.ctx);
+	const signed = readFileSync(f.path, "utf8");
+	if (failure === "empty") writeFileSync(f.path, "");
+	else { rmSync(f.path); if (failure === "directory") mkdirSync(f.path); }
+	await delay(250); // also exercise unavailable read after debounce has expired
+	f.hooks.get("agent_end")({}, f.ctx);
+	expect(f.entries.at(-1).data.signoffs["first output"]).toBeDefined();
+	f.hooks.get("session_compact")();
+	const unavailable = f.hooks.get("before_agent_start")({ systemPrompt: "base" }, f.ctx);
+	expect(unavailable.systemPrompt).toContain("unavailable");
+	expect(unavailable.message).toBeUndefined();
+	if (failure === "directory") rmSync(f.path, { recursive: true });
+	writeFileSync(f.path, signed);
+	const resync = f.hooks.get("before_agent_start")({ systemPrompt: "base" }, f.ctx);
+	expect(resync.message.content).toContain("Observed PASS");
+	await delay(250);
+	expect(f.entries.at(-1).data.signoffs["first output"]).toBeDefined();
+	expect(f.changed()).toBe(0);
+});
+
+it("ignores post-completion maintenance but reviews evidence, requirement or manual reopening changes", async () => {
+	const f = fixture(); await f.draft(); await f.command("ready");
+	writeFileSync(join(f.ctx.cwd, "proof.log"), "PASS\n");
+	for (const goal of ["first output", "second output"]) await f.tools.get("CompleteGoal").execute("c", { goal, evidence: ["proof.log"], observation: "PASS" }, undefined, undefined, f.ctx);
+	const signed = readFileSync(f.path, "utf8");
+	await f.atomicWrite(signed.replace("## Log", "## Log\n- recap: finished"));
+	await delay(250);
+	expect(f.changed()).toBe(0); // Log-only edits are history, not requirements
+	// Worker-authored evidence above the Log must surface: a supervisor caught a worker's
+	// contradictory evidence block through exactly this event (LUCID3, 2026-09-10).
+	await f.atomicWrite(signed.replace("## Log", "  - evidence: proof.log\n## Log\n- recap: finished"));
+	await waitFor(() => f.changed() === 1);
+	await f.atomicWrite(signed.replace("## Log", "- discriminator: exact bytes and trailing newline\n## Log"));
+	await waitFor(() => f.changed() === 2);
+	await f.atomicWrite(signed.replace("[x] goal: first", "[ ] goal: first"));
+	await waitFor(() => f.changed() === 3);
+	expect(f.entries.at(-1).data.signoffs["first output"]).toBeUndefined();
+});
+
+it("cancelled goals do not prevent final cleanup, and solo writes self-verification in Log", async () => {
+	const f = fixture(); await f.draft();
+	writeFileSync(f.path, f.plan.replace("[ ] goal: second", "[-] goal: second") + "\n## Appendix\nPreserved context\n");
+	f.ctx.ui.select.mockResolvedValueOnce("Worker confirmed stopped"); await f.command("solo");
+	writeFileSync(join(f.ctx.cwd, "proof.log"), "PASS\n");
+	const done = await f.tools.get("CompleteGoal").execute("c", { goal: "first output", evidence: ["proof.log"], observation: "Exact bytes observed" }, undefined, undefined, f.ctx);
+	expect(done.content[0].text).toContain("All non-cancelled goals are reviewed");
+	const text = readFileSync(f.path, "utf8");
+	expect(text).toContain("Solo self-verification:");
+	expect(text).not.toContain("Parent review:");
+	expect(text.indexOf("Solo self-verification:")).toBeLessThan(text.indexOf("## Appendix"));
+	expect(text).toContain("Preserved context");
+});
+
+it("lineage-only child attaches its plan with goal-only widget, retains task context, and cannot complete", async () => {
+	const f = fixture(true);
+	const supplied = join(f.ctx.cwd, "supplied.md");
+	const text = "- [/] goal: exact file\n  - [ ] verify bytes\n## Log\n  - [ ] archived task\n";
+	writeFileSync(supplied, text);
+	const before = f.hooks.get("before_agent_start")({ systemPrompt: "base" }, f.ctx);
+	expect(before.systemPrompt).toContain("AttachGoalPlan");
+	const attach = f.tools.get("AttachGoalPlan");
+	await attach.execute("a", { path: "supplied.md" }, undefined, undefined, f.ctx);
+	expect(f.entries.at(-1).data.plan).toBeUndefined(); // no cwd heuristics
+	await attach.execute("a", { path: supplied }, undefined, undefined, f.ctx);
+	expect(f.ctx.ui.setWidget.mock.lastCall?.[1]).toEqual(["▸ exact file"]);
+	expect(f.ctx.ui.setWidget.mock.lastCall?.[1].join("\n")).not.toContain("archived task");
+	expect(readFileSync(supplied, "utf8")).toBe(text);
+	f.hooks.get("session_start")({}, f.ctx);
+	expect(f.hooks.get("before_agent_start")({ systemPrompt: "base" }, f.ctx).message.content).toContain("exact file");
+	const completion = await f.tools.get("CompleteGoal").execute("c", { goal: "exact file", evidence: [], observation: "claim" }, undefined, undefined, f.ctx);
+	expect(completion.content[0].text).toContain("only to the active parent");
+});
+
+it.each(["solo", "supervising"])("%s widget omits long tasks without altering the plan", async mode => {
+	const f = fixture(); await f.draft();
+	const text = "- [/] goal: first output\n  - [ ] a long task that should never take widget space\n- [ ] goal: second output\n## Log\n";
+	writeFileSync(f.path, text);
+	if (mode === "solo") { f.ctx.ui.select.mockResolvedValueOnce("Worker confirmed stopped"); await f.command("solo"); }
+	else await f.command("ready");
+	expect(f.ctx.ui.setWidget.mock.lastCall?.[1]).toEqual(["▸ first output", "○ second output"]);
+	expect(readFileSync(f.path, "utf8")).toBe(text);
+});
+
+it.each(["solo", "supervising"])("%s upkeep is turn-driven, folds Log, resets on working-set edits, and never starts a turn", async mode => {
+	const f = fixture(); await f.draft();
+	if (mode === "solo") { f.ctx.ui.select.mockResolvedValueOnce("Worker confirmed stopped"); await f.command("solo"); }
+	else await f.command("ready");
+	f.hooks.get("before_agent_start")({ systemPrompt: "base" }, f.ctx);
+	const reminders = () => f.messages.filter(m => m.message.customType === "pi-goals-upkeep");
+	f.hooks.get("turn_end")({}, f.ctx); // observe initial working set
+	for (let i = 0; i < 7; i++) {
+		writeFileSync(f.path, f.plan + `- historical recap ${i}\n`);
+		f.hooks.get("turn_end")({}, f.ctx);
+	}
+	expect(reminders()).toHaveLength(0);
+	f.hooks.get("turn_end")({}, f.ctx);
+	expect(reminders()).toHaveLength(1);
+	expect(reminders()[0].options).toEqual({ triggerTurn: false });
+	expect(reminders()[0].message.content).toContain(f.path);
+	expect(reminders()[0].message.content).not.toContain("first output");
+	for (let i = 0; i < 16; i++) f.hooks.get("turn_end")({}, f.ctx);
+	expect(reminders()).toHaveLength(1);
+	expect(reminders()[0].message.content).not.toContain("historical recap");
+	for (let i = 0; i < 7; i++) f.hooks.get("turn_end")({}, f.ctx);
+	writeFileSync(f.path, f.plan.replace("first output", "refined output"));
+	f.hooks.get("turn_end")({}, f.ctx);
+	expect(reminders()).toHaveLength(1);
+	await f.command("stop");
+	for (let i = 0; i < 10; i++) f.hooks.get("turn_end")({}, f.ctx);
+	expect(reminders()).toHaveLength(1);
+});
+
+it("extra subagent launches are recorded as helpers and never steal the implementation identity", async () => {
+	const f = fixture(); await f.draft(); await f.command("ready");
+	f.hooks.get("tool_result")({ toolName: "subagent", details: { id: "impl", sessionFile: "/tmp/impl.jsonl" } });
+	expect(f.entries.at(-1).data).toMatchObject({ worker: { id: "impl", sessionFile: "/tmp/impl.jsonl" }, helpers: [] });
+	f.hooks.get("tool_result")({ toolName: "subagent", details: { id: "reviewer", sessionFile: "/tmp/review.jsonl" } });
+	expect(f.entries.at(-1).data).toMatchObject({ worker: { id: "impl" }, helpers: [{ id: "reviewer", sessionFile: "/tmp/review.jsonl" }] });
+	// a repeated helper launch updates its record instead of duplicating it
+	f.hooks.get("tool_result")({ toolName: "subagent", details: { id: "reviewer-2", sessionFile: "/tmp/review.jsonl" } });
+	expect(f.entries.at(-1).data.helpers).toEqual([{ id: "reviewer-2", sessionFile: "/tmp/review.jsonl" }]);
+	// resuming the worker keeps the binding and refreshes its id
+	f.hooks.get("tool_result")({ toolName: "subagent_resume", details: { id: "impl-2", sessionFile: "/tmp/impl.jsonl" } });
+	expect(f.entries.at(-1).data).toMatchObject({ worker: { id: "impl-2", sessionFile: "/tmp/impl.jsonl" }, helpers: [{ id: "reviewer-2" }] });
+});
+
+it("pending launch counter survives concurrent launches until every result lands", async () => {
+	const f = fixture(); await f.draft(); await f.command("ready");
+	f.hooks.get("tool_call")({ toolName: "subagent" });
+	f.hooks.get("tool_call")({ toolName: "subagent" });
+	f.hooks.get("tool_result")({ toolName: "subagent", details: { id: "a", sessionFile: "/tmp/a.jsonl" } });
+	f.ctx.ui.select.mockResolvedValueOnce("Worker confirmed stopped");
+	await f.command("solo");
+	expect(f.entries.at(-1).data.mode).toBe("supervising"); // one launch still pending
+	expect(f.ctx.notify ?? f.ctx.ui.notify).toHaveBeenLastCalledWith(expect.stringContaining("still pending"), "warning");
+	f.hooks.get("tool_result")({ toolName: "subagent", details: { id: "b", sessionFile: "/tmp/b.jsonl" } });
+	f.ctx.ui.select.mockResolvedValueOnce("Worker confirmed stopped");
+	await f.command("solo");
+	expect(f.entries.at(-1).data.mode).toBe("solo");
+	expect(f.entries.at(-1).data).toMatchObject({ worker: { id: "a" }, helpers: [{ id: "b" }] });
+});
+
+it("late worker results invalidate a takeover menu but do not disable plan watching", async () => {
+	const f = fixture(); await f.draft(); await f.command("ready");
+	let answer!: (choice: string) => void;
+	f.ctx.ui.select.mockImplementationOnce(() => new Promise(resolve => { answer = resolve; }));
+	const solo = f.command("solo");
+	f.hooks.get("tool_result")({ toolName: "subagent", details: { id: "late-child", sessionFile: "/tmp/late.jsonl" } });
+	answer("Worker confirmed stopped"); await solo;
+	expect(f.entries.at(-1).data.mode).toBe("supervising");
+	expect(f.entries.at(-1).data.workerStopped).toBe(false);
+	await f.atomicWrite(f.plan.replace("first output", "new requirement"));
+	await waitFor(() => f.changed() === 1);
+});
+
+it("changed plan or shutdown during takeover never grants solo permission", async () => {
+	const f = fixture(); await f.draft();
+	f.ctx.ui.select.mockImplementationOnce(async () => { writeFileSync(f.path, f.plan.replace("first", "changed")); return "Worker confirmed stopped"; });
+	await f.command("solo");
+	expect(f.entries.at(-1).data.mode).toBe("planning");
+	expect(f.ctx.ui.notify).toHaveBeenLastCalledWith(expect.stringContaining("Plan changed during takeover"), "warning");
+	f.ctx.ui.select.mockImplementationOnce(async () => { f.shutdown(); return "Worker confirmed stopped"; });
+	await f.command("solo");
+	expect(f.entries.at(-1).data.mode).toBe("planning");
+});
+
+it("requires explicit supervisor ownership confirmation when attaching an existing plan", async () => {
+	const f = fixture(); const path = join(f.ctx.cwd, "shared.md");
+	writeFileSync(path, f.plan);
+	f.ctx.ui.select.mockResolvedValueOnce("Cancel");
+	await f.command(`attach ${path}`);
+	expect(f.entries).toHaveLength(0);
+	f.ctx.ui.select.mockResolvedValueOnce("Previous supervisor confirmed stopped");
+	await f.command(`attach ${path}`);
+	expect(f.entries.at(-1).data).toMatchObject({ mode: "planning", plan: path });
+});
+
+it("does not approve cancelled goals or display current completion for an unavailable plan", async () => {
+	const f = fixture(); await f.draft(); await f.command("ready");
+	writeFileSync(f.path, "- [-] goal: cancelled output\n## Log\n");
+	writeFileSync(join(f.ctx.cwd, "evidence.log"), "verified\n");
+	const reply = await f.tools.get("CompleteGoal").execute("t", { goal: "cancelled output", evidence: ["evidence.log"], observation: "read" }, undefined, undefined, f.ctx);
+	expect(reply.content[0].text).toContain("no sign-off recorded");
+	expect(readFileSync(f.path, "utf8")).toContain("[-]");
+	rmSync(f.path);
+	f.hooks.get("agent_end")({}, f.ctx);
+	expect(f.ctx.ui.setWidget).toHaveBeenLastCalledWith("goals", [expect.stringContaining("unavailable")]);
+});
+
+it("uses scheduler storage for ownership and the real public user controls", () => {
+	const prompt = scheduleCheckIn("session-1", "/plan.md");
+	expect(prompt).toContain(".pi/schedule-prompts.json");
+	expect(prompt).toContain("tool text does not expose binding");
+	expect(prompt).toContain("Never use cleanup");
+	expect(prompt).toContain("deletes disabled jobs");
+	expect(prompt).toContain("schedule_prompt update");
+	expect(prompt).not.toContain("with /schedule-prompts");
+});
+
+it("keeps interactive workers open and supplies the supervisor identity for Intercom reports", async () => {
+	const agent = readFileSync(new URL("../agents/goals-worker.md", import.meta.url), "utf8");
+	expect(agent).toContain("auto-exit: false");
+	const f = fixture(); await f.draft(); await f.command("ready");
+	expect(f.messages.at(-1).message.content).toContain("supervisor Intercom session copy-only");
+	const role = f.hooks.get("before_agent_start")({ systemPrompt: "base" }, f.ctx).systemPrompt;
+	expect(role).toContain("your Intercom session ID copy-only");
+	expect(role).toContain("stop workers before /reload");
+	expect(role).not.toContain("Reports arrive automatically");
+});

test/judge-args.test.ts

@@ -1,33 +0,0 @@
-import { describe, expect, it } from "vitest";
-import { buildJudgeArgs } from "../src/index.js";
-
-describe("buildJudgeArgs", () => {
-	it("omits --model when judgeModel is null (pi uses its configured default; never a pre-emptive 'no model' failure)", () => {
-		const args = buildJudgeArgs(null);
-		expect(args).not.toContain("--model");
-		// an empty --model "" would make every sign-off silently inconclusive -- guard against it
-		const i = args.indexOf("--model");
-		expect(i).toBe(-1);
-	});
-
-	it("includes --model <ref> when an explicit/session model is set", () => {
-		const args = buildJudgeArgs("openrouter/~anthropic/claude-haiku-latest");
-		const i = args.indexOf("--model");
-		expect(i).not.toBe(-1);
-		expect(args[i + 1]).toBe("openrouter/~anthropic/claude-haiku-latest");
-	});
-
-	it("always sets --no-session, --no-extensions, the read-only tool allowlist, and edit/write exclusion", () => {
-		for (const m of [null, "some/model"]) {
-			const args = buildJudgeArgs(m);
-			expect(args).toContain("--no-session");
-			expect(args).toContain("--no-extensions"); // a broken global extension must not take down sign-offs
-			expect(args).toContain("--tools");
-			expect(args.some((a) => a.startsWith("read,grep,find,ls"))).toBe(true);
-			// no bash: the judge must never be able to execute (or re-run a 10-hour verify) or mutate
-			expect(args.some((a) => a.includes("bash"))).toBe(false);
-			expect(args).toContain("--exclude-tools");
-			expect(args.some((a) => a.includes("edit") && a.includes("write"))).toBe(true);
-		}
-	});
-});

test/package.test.ts

@@ -0,0 +1,15 @@
+import { existsSync, readFileSync } from "node:fs";
+import { resolve } from "node:path";
+import { expect, it } from "vitest";
+
+it("declares current entry and bundled extension resources that exist after install", () => {
+	const manifest = JSON.parse(readFileSync("package.json", "utf8"));
+	expect(manifest.pi.extensions[0]).toBe("./src/index.ts");
+	for (const path of manifest.pi.extensions) expect(existsSync(resolve(path)), path).toBe(true);
+	for (const name of ["pi-subagents", "pi-intercom", "pi-schedule-prompt"]) {
+		expect(manifest.dependencies[name]).toBeTruthy();
+		expect(manifest.bundledDependencies).toContain(name);
+	}
+	expect(manifest.dependencies["pi-subagents"]).toContain("953c6f6d2fc7d8a5c956c30cd77c51bad697c2a4");
+	expect(existsSync("agents/goals-worker.md")).toBe(true);
+});

test/plan-view.test.ts

@@ -0,0 +1,40 @@
+import { expect, it } from "vitest";
+import { planViews } from "../src/plan-view.js";
+
+it("keeps outcome, preferences and discriminators without tasks or history", () => {
+	const plan = "# Outcome\nBeat random, not just plot it.\n## User preferences\nKeep costs low.\n## Goals\n1. [ ] goal: repair\n  - discriminator: beats random\n  - subtle failure mode: plot exists but result fails\n  - tasks:\n    1. [x] draw plot\n  - evidence:\n    - old output\n2. [ ] goal: confirm\n## Task list\n- [ ] run it\n## Appendix\nunapproved idea";
+	const views = planViews(plan);
+	for (const text of ["Beat random", "Keep costs low", "goal: repair", "discriminator: beats random", "subtle failure mode", "goal: confirm"]) expect(views.short).toContain(text);
+	for (const text of ["draw plot", "old output", "run it", "unapproved idea"]) expect(views.short).not.toContain(text);
+	expect(views.long).toContain("draw plot");
+	expect(views.long).toContain("old output");
+	expect(views.long).not.toContain("unapproved idea");
+});
+
+it("omits only named worker identity fields from review while retaining them in full context", () => {
+	const base = "# Plan\n- preferred worker model: provider/model\n- [ ] goal: result\n  - discriminator: exact bytes";
+	const metadata = "\n- Active worker: worker-1\n- worker session: /saved.jsonl\n- worker intercom session: uuid";
+	expect(planViews(base + metadata).short).toBe(planViews(base).short);
+	expect(planViews(base + metadata).long).toContain("/saved.jsonl");
+	expect(planViews(base.replace("exact bytes", "approximate match")).short).not.toBe(planViews(base).short);
+	expect(planViews(base.replace("[ ]", "[x]")).short).not.toBe(planViews(base).short);
+});
+
+it("notifies on goal and task changes but not on identity bookkeeping or log edits", () => {
+	const base = "# Plan\n- [ ] goal: result\n## Task list\n- [ ] run it\n- worker session: /saved.jsonl\n## Log\nfirst entry";
+	const baseView = planViews(base).notify;
+	// identity bookkeeping: silent
+	expect(planViews(base.replace("/saved.jsonl", "/moved.jsonl")).notify).toBe(baseView);
+	// log edits: silent
+	expect(planViews(base.replace("first entry", "second entry")).notify).toBe(baseView);
+	// worker ticking a task: review event (field catch, LUCID3 2026-09-10)
+	expect(planViews(base.replace("- [ ] run it", "- [x] run it")).notify).not.toBe(baseView);
+	// goal edits: review event
+	expect(planViews(base.replace("[ ] goal: result", "[x] goal: result")).notify).not.toBe(baseView);
+});
+
+it("stops at history and preserves a manual goal tick", () => {
+	const view = planViews("# Plan\n1. [x] goal: result\n## Log\n1. [ ] goal: historical");
+	expect(view.short).toContain("[x] goal: result");
+	expect(view.long).not.toContain("historical");
+});

test/prompts.test.ts

@@ -0,0 +1,21 @@
+import { describe, expect, it } from "vitest";
+import { planDrafting } from "../src/prompts.js";
+
+describe("planning prompt", () => {
+	it("requires fact finding or a focused question before a goal", () => {
+		expect(planDrafting).toContain("Use read-only repository tools or web search when either can\nresolve a fact.");
+		expect(planDrafting).toContain("Do not use a question quota");
+		expect(planDrafting).toContain("Briefly reframe the request in your own words to check comprehension");
+		expect(planDrafting).toContain("point as unknown; do not silently replace it with an inference or turn it into a new blocking decision");
+		expect(planDrafting).toContain("answer materially reduces uncertainty\nwhile discovering the right plan");
+		expect(planDrafting).toContain("self-contained: state the relevant\ncontext, use the human's language and ASD-STE100");
+		expect(planDrafting).toContain("placeholder goal such as \"work out the thing\"");
+		expect(planDrafting).toContain("Only withhold Ready for an unanswered choice that changes scope, spending, or the user-visible result");
+	});
+
+	it("anchors work and sign-off to the user-visible result", () => {
+		expect(planDrafting).toContain("## User-visible result");
+		expect(planDrafting).toContain("Take it from the original request, not from your implementation plan");
+		expect(planDrafting).toContain("Future work may not defer any artifact or action named there");
+	});
+});

test/rpc-review.test.ts

@@ -0,0 +1,166 @@
+import { type ChildProcessWithoutNullStreams, spawn } from "node:child_process";
+import { once } from "node:events";
+import { mkdtempSync, readFileSync, rmSync } from "node:fs";
+import { createServer } from "node:http";
+import { tmpdir } from "node:os";
+import { join, resolve } from "node:path";
+import { StringDecoder } from "node:string_decoder";
+import { describe, expect, it } from "vitest";
+
+type RpcMessage = { type: string; id?: string; method?: string; [key: string]: unknown };
+type ModelRequest = { messages: Array<{ role: string; content: unknown }> };
+
+class RpcClient {
+	readonly messages: RpcMessage[] = [];
+	stderr = "";
+	private readonly waiters: Array<{ predicate: (message: RpcMessage) => boolean; resolve: (message: RpcMessage) => void }> = [];
+
+	constructor(readonly process: ChildProcessWithoutNullStreams) {
+		const decoder = new StringDecoder("utf8");
+		let buffer = "";
+		process.stderr.on("data", (chunk) => { this.stderr += chunk; });
+		process.stdout.on("data", (chunk) => {
+			buffer += decoder.write(chunk);
+			while (buffer.includes("\n")) {
+				const newline = buffer.indexOf("\n");
+				const line = buffer.slice(0, newline).replace(/\r$/, "");
+				buffer = buffer.slice(newline + 1);
+				if (!line) continue;
+				const message = JSON.parse(line) as RpcMessage;
+				this.messages.push(message);
+				const index = this.waiters.findIndex(({ predicate }) => predicate(message));
+				if (index !== -1) this.waiters.splice(index, 1)[0].resolve(message);
+			}
+		});
+	}
+
+	send(message: RpcMessage): void {
+		this.process.stdin.write(`${JSON.stringify(message)}\n`);
+	}
+
+	waitFor(predicate: (message: RpcMessage) => boolean, after = 0): Promise<RpcMessage> {
+		const existing = this.messages.slice(after).find(predicate);
+		if (existing) return Promise.resolve(existing);
+		return new Promise((resolvePromise, reject) => {
+			const timer = setTimeout(() => {
+				this.waiters.splice(this.waiters.indexOf(waiter), 1);
+				reject(new Error(`RPC wait timed out: ${this.stderr}\n${JSON.stringify(this.messages.slice(-12))}`));
+			}, 8_000);
+			const waiter = { predicate, resolve: (message: RpcMessage) => { clearTimeout(timer); resolvePromise(message); } };
+			this.waiters.push(waiter);
+		});
+	}
+}
+
+function streamResponse(response: import("node:http").ServerResponse, delta: object, finishReason: "stop" | "tool_calls"): void {
+	response.writeHead(200, { "content-type": "text/event-stream" });
+	response.write(`data: ${JSON.stringify({ choices: [{ index: 0, delta, finish_reason: null }] })}\n\n`);
+	response.write(`data: ${JSON.stringify({ choices: [{ index: 0, delta: {}, finish_reason: finishReason }] })}\n\n`);
+	response.end("data: [DONE]\n\n");
+}
+
+const isSelect = (message: RpcMessage) => message.type === "extension_ui_request" && message.method === "select";
+const isEditor = (message: RpcMessage) => message.type === "extension_ui_request" && message.method === "editor";
+const systemText = (request: ModelRequest) => request.messages.filter(message => ["system", "developer"].includes(message.role)).map(message => message.content).join("\n");
+
+describe("RPC review flow", () => {
+	it.each(["Edit", "Discuss"])("automatically proposes a draft, handles %s, then enters the supervisor role on Ready", async (choice) => {
+		const cwd = mkdtempSync(join(tmpdir(), "pi-goals-rpc-"));
+		const requests: ModelRequest[] = [];
+		const plan = "# Plan\n\n## Goals\n\n1. [ ] goal: name the output\n  - subtle failure mode: the output has no name\n  - discriminator: the plan names the output\n\n## Log\n";
+		let planPath = "";
+		const server = createServer(async (request, response) => {
+			let body = "";
+			for await (const chunk of request) body += chunk;
+			requests.push(JSON.parse(body));
+			if (requests.length === 1) {
+				streamResponse(response, {
+					tool_calls: [{
+						index: 0, id: "write-plan", type: "function",
+						function: { name: "write", arguments: JSON.stringify({ path: planPath, content: plan }) },
+					}],
+				}, "tool_calls");
+				return;
+			}
+			streamResponse(response, { content: "Plan inspected." }, "stop");
+		});
+		await new Promise<void>((done) => server.listen(0, "127.0.0.1", done));
+		const address = server.address();
+		if (!address || typeof address === "string") throw new Error("Offline model did not bind a TCP port.");
+
+		const pi = spawn(resolve("node_modules/.bin/pi"), [
+			"--mode", "rpc", "--no-session", "--no-extensions", "--model", "offline/test",
+			"-e", resolve("test/fixtures/offline-model.ts"),
+			"-e", resolve("test/fixtures/subagent-schema.ts"),
+			"-e", resolve("src/index.ts"),
+		], {
+			cwd,
+			env: {
+				// Pi/gpt-6-astra: test the parent role even when vitest itself runs in a worker.
+				...Object.fromEntries(Object.entries(process.env).filter(([name]) => !name.startsWith("PI_SUBAGENT_") && !name.startsWith("PI_GOALS_"))),
+				PI_CODING_AGENT_DIR: join(cwd, ".agent"),
+				PI_GOALS_OFFLINE_MODEL_URL: `http://127.0.0.1:${address.port}`,
+			},
+		});
+		const client = new RpcClient(pi);
+		const exited = once(pi, "exit");
+		try {
+			client.send({ type: "get_state", id: "state" });
+			const state = await client.waitFor((message) => message.type === "response" && message.id === "state");
+			const sessionId = (state.data as { sessionId: string }).sessionId;
+			planPath = join(cwd, ".pi", "plan", `${sessionId}-main.md`);
+
+			client.send({ type: "prompt", id: "goals", message: "/goals new work out the thing" });
+			const review = await client.waitFor(isSelect);
+			expect(review.options).toEqual(["Ready", "Discuss", "Edit", "Cancel"]);
+			expect(review.title).toContain(planPath);
+			const proposal = client.messages.find(message => message.type === "message_end" && (message.message as { customType?: string })?.customType === "goal-plan-proposal");
+			expect(proposal?.message).toMatchObject({ content: plan, display: true });
+			expect(readFileSync(planPath, "utf8")).toBe(plan);
+			expect(requests).toHaveLength(2);
+			expect(systemText(requests[0])).toContain("Plan only in");
+
+			const choiceStart = client.messages.length;
+			client.send({ type: "extension_ui_response", id: review.id, value: choice });
+			let approvedPlan = plan;
+			if (choice === "Edit") {
+				const editor = await client.waitFor(isEditor, choiceStart);
+				expect(editor.prefill).toBe(plan);
+				expect(requests).toHaveLength(2);
+				approvedPlan = plan.replace("the plan names the output", "the plan names output.txt and its exact bytes");
+				const editStart = client.messages.length;
+				client.send({ type: "extension_ui_response", id: editor.id, value: approvedPlan });
+				await client.waitFor(message => message.type === "extension_ui_request" && message.method === "setWidget", editStart);
+				expect(readFileSync(planPath, "utf8")).toBe(approvedPlan);
+				expect(requests).toHaveLength(2);
+			} else {
+				await client.waitFor(message => message.type === "agent_end", choiceStart);
+				expect(requests).toHaveLength(3);
+				expect(systemText(requests[2])).toContain("Plan only in");
+				expect(JSON.stringify(requests[2].messages.at(-1))).toContain("Discuss the current draft");
+				expect(client.messages.slice(choiceStart).filter(isEditor)).toEqual([]);
+			}
+			const beforeReady = requests.length;
+			const reopenStart = client.messages.length;
+			client.send({ type: "prompt", id: "review", message: "/goals review" });
+			const ready = await client.waitFor(isSelect, reopenStart);
+			expect(requests).toHaveLength(beforeReady);
+			const readyStart = client.messages.length;
+			client.send({ type: "extension_ui_response", id: ready.id, value: "Ready" });
+			await client.waitFor(message => message.type === "agent_end", readyStart);
+			expect(requests).toHaveLength(beforeReady + 1);
+			const supervisor = requests.at(-1)!;
+			expect(systemText(supervisor)).toContain("You are the goal supervisor in the main chat");
+			expect(systemText(supervisor)).not.toContain("Plan only in");
+			expect(JSON.stringify(supervisor.messages)).toContain(JSON.stringify(approvedPlan).slice(1, -1));
+			expect(client.messages.filter(message => message.type === "tool_execution_start").map(message => message.toolName)).toEqual(["write"]);
+			expect(client.messages.filter(message => message.type === "extension_error")).toEqual([]);
+			console.log(`RPC ${choice}: visible automatic proposal; ${choice === "Edit" ? "editor saved exact plan without model call" : "discussion retained planning role without editor"}; Ready request used supervisor role; only write executed.`);
+		} finally {
+			pi.kill();
+			await exited;
+			await new Promise<void>((done) => server.close(() => done()));
+			rmSync(cwd, { recursive: true, force: true });
+		}
+	}, 25_000);
+});

test/session-usage.test.ts

@@ -0,0 +1,34 @@
+import { expect, it } from "vitest";
+import { report, summarize } from "../scripts/session-usage.mjs";
+
+const start = "2026-09-10T06:00:00.000Z";
+const end = "2026-09-10T07:00:00.000Z";
+const request = (timestamp: string, model = "a") => ({
+	type: "message", timestamp, message: { role: "assistant", provider: "test", model,
+		usage: { input: 10, cacheRead: 100, cacheWrite: 5, output: 20, reasoning: 8, totalTokens: 135 } },
+});
+
+it("excludes inherited history and counts repeated cached input without adding reasoning twice", () => {
+	const result = summarize([request("2026-09-09T06:00:00.000Z"), request(start), request(end, "b")], start, end);
+	expect(result).toMatchObject({ calls: 2, input: 20, cacheRead: 200, cacheWrite: 10, output: 40, totalTokens: 270 });
+	expect(result.models.map((m: any) => m.model)).toEqual(["test/a", "test/b"]);
+});
+
+it("uses the latest planning start and the same interval for both sessions", () => {
+	const supervisor = { entries: [
+		{ type: "custom", customType: "pi-goals-main-supervisor-v1", timestamp: start, id: "boundary", data: { mode: "planning", plan: "plan.md" } },
+		request(start),
+	] };
+	const result = report(supervisor, { entries: [request(end)] }, end);
+	expect(result.since).toBe(start);
+	expect(result.elapsedHours).toBe(1);
+	expect(result.sessions.map((s: any) => s.output)).toEqual([20, 20]);
+	expect(() => report({ entries: [] }, { entries: [] }, end)).toThrow("No recorded planning start");
+});
+
+it("reports missing usage and rejects invalid recorded token counts", () => {
+	const missing = { type: "message", timestamp: start, message: { role: "assistant" } };
+	expect(summarize([missing], start, end).missingUsage).toBe(1);
+	const invalid = request(start); invalid.message.usage.input = Number.NaN;
+	expect(() => summarize([invalid], start, end)).toThrow("Invalid usage.input");
+});

test/tick-goal.test.ts

@@ -1,32 +0,0 @@
-import { describe, expect, it } from "vitest";
-import { tickGoal } from "../src/index.js";
-
-const plan = `# Plan
-
-## Goals
-
-1. [/] goal: Implement the cache layer
-  - tasks:
-    1. [x] wire client
-2. [ ] goal: Ship the docs
-
-## Log
-`;
-
-describe("tickGoal (sign-off ticks the goal; agent only ticks on wording drift)", () => {
-	it("ticks the exact-matching goal line, case-insensitive, leaving subtasks alone", () => {
-		const out = tickGoal(plan, "implement the CACHE layer");
-		expect(out).toContain("1. [x] goal: Implement the cache layer");
-		expect(out).toContain("1. [x] wire client"); // subtask untouched (was already x)
-		expect(out).toContain("2. [ ] goal: Ship the docs"); // other goal untouched
-	});
-
-	it("returns null on wording drift (fuzzy matching is the judge's job, not TypeScript's)", () => {
-		expect(tickGoal(plan, "Implement caching")).toBeNull();
-	});
-
-	it("returns null when the subject matches more than one goal line", () => {
-		const dup = `${plan}3. [ ] goal: Ship the docs\n`;
-		expect(tickGoal(dup, "Ship the docs")).toBeNull();
-	});
-});

tsconfig.json

@@ -6,10 +6,12 @@
     "strict": true,
     "esModuleInterop": true,
     "skipLibCheck": true,
-    "jsx": "react-jsx",
     "outDir": "dist",
     "rootDir": "src",
     "declaration": true
   },
-  "include": ["src/**/*.ts", "src/**/*.tsx"]
+  "include": [
+    "src/**/*.ts",
+    "src/**/*.tsx"
+  ]
 }