pi-plan

A pi extension for plan-driven, goal-tracked work in one plan.md. Set up goals (with evidence and failure modes) in plan mode, work them, and sign a goal off only when a read-only subagent has checked the evidence.

Successor to pi-lgtm, kept deliberately small: about burneikis/pi-plan plus the additions, goals with evidence, a sign-off check, a widget, and a reminder.

The form guides; it does not gate. The agent edits plan.md with its normal Edit tool. The one blessed tool is CompleteGoal, which runs the sign-off check and records the result. The reminder, the injected plan summary, and git/widget visibility carry the process. It trusts the agent's judgement rather than guarding it.

Install

pi install npm:@wassname2/pi-plan

Or run without installing:

pi -e npm:@wassname2/pi-plan

Use

/plan add CSV export to the report view

1. Plan. The agent explores read-only and writes goals into plan.md (see format below).
2. Review. You get a menu: Ready, Edit (ask the agent to revise), Open in $EDITOR, or Cancel. On Ready you choose whether to keep the current context or start fresh and compacted.
3. Work. Each turn the active goal is injected (so it survives compaction) and a reminder nudges the agent to keep plan.md current and work autonomously. When a goal's done_when is met the agent calls CompleteGoal, which runs verify and a read-only judge and, on accept, marks it done and logs it.

Other commands: /plan (print the plan), /plan clear (empty plan.md, history kept in git), /plan judge <model-ref> (use a specific model for the sign-off judge; default is your current model).

plan.md format

One file holds the objective, the goals, and a short append-only log.

# Plan: ship the cache layer

## 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
- [x] wire cache client
- [ ] eviction policy

## Log
- 2026-06-15 14:02  cache client wired; eviction next

• A goal is a ## Goal: header with an <!-- id -->, a status: (open | active | done | cancelled), a falsifiable done_when: (what you expect, and the symptom if it is NOT met), an optional verify: shell command, a failure_modes: pre-mortem list, and - [ ] subtasks.
• done_when names the evidence that distinguishes real success from a subtle failure. verify, when present, is the deterministic first stage of the sign-off check.
• The agent ticks subtasks, appends to ## Log, and sets status as it works. Multiple goals may be active.

The sign-off check (CompleteGoal)

CompleteGoal(goal_id, evidence, paths?) is the one blessed completion path:

1. If the goal has a verify: command, it is run. A non-zero exit rejects immediately, with no model call.
2. Otherwise a read-only pi subprocess (the judge) inspects the evidence against the repo and the named failure modes and returns a verdict. It re-derives from the artifacts you point it at rather than trusting the claim, so point evidence/paths at durable artifacts (saved logs, committed diffs, files).
3. On accept, the goal's status flips to done and a ## Log line is written. On reject, the goal stays open and the agent is told what is missing.

The judge defaults to your current model (guaranteed authorized and capable). Set a different one with /plan judge <provider/model> for an independent cross-family check.

Prompts

All model-facing text lives in src/prompts.ts, in flow order, so the process is easy to review end to end.

Develop

pi -e ./src/index.ts        # load locally
npm test                    # vitest: parser + sign-off record logic
npm run typecheck
npm run lint

Not (yet) included

No autonomous re-prompt loop (an until-done-style loop judge). Autonomy comes from the reminder, not a harness. Plan-phase model stickiness is a documented next step.

License

MIT