Planning blocked every subagent launch, so oracles and scouts could not run before Ready. Now only OpenGoalWorker and project.* are blocked in planning; pause still blocks all launches.
Co-Authored-By: PI[gpt-6-sol] <288921227+claudypoo@users.noreply.github.com>
Removes ~1850 words of unreviewed agent prose added since 21f1092; keeps the approved check-in paragraph and status legend. AGENTS.md: README changes need sign-off.
Co-Authored-By: PI[gpt-6-sol] <288921227+claudypoo@users.noreply.github.com>
Ground each recommendation and optionally predict the user's preference with evidence and confidence; keep the forecast distinct from the recommendation.\n\nCo-Authored-By: PI[gpt-6-sol] <288921227+claudypoo@users.noreply.github.com>
Replace direct scheduler and VCC tarball specs with exact registry versions. Keep the existing lockfile integrity and bundled extension resources.\n\nCo-Authored-By: PI[gpt-6-sol] <288921227+claudypoo@users.noreply.github.com>
Append a UI-only saved entry after supervisor or solo final updates; omit short unchanged waits. Remove the prompt instructions to sample fortunes, and cover the non-context renderer and lifecycle.
Co-Authored-By: PI/OpenAI <288921227+claudypoo@users.noreply.github.com>
Retain the user-visible result in compact reminders, replace bookkeeping-first instructions, and remove a brittle wording assertion. Record observed supervision failures and document local-path development outside Pi's managed Git cache.
Co-Authored-By: PI/OpenAI <288921227+claudypoo@users.noreply.github.com>
Let OpenGoalWorker defer pane concurrency to stock controls. Preserve old bindings when stock reports an existing pane; supersede them only when a replacement pane is newly opened. No human or model attestation gate.\n\nCo-Authored-By: PI/OpenAI <288921227+claudypoo@users.noreply.github.com>
Remove the infrastructure confirmation modal. Require an auditable model observation, record the released binding, and continue through OpenGoalWorker without asking the human to diagnose panes or transport state.\n\nCo-Authored-By: PI/OpenAI <288921227+claudypoo@users.noreply.github.com>
Render routine wake-ups as one expandable custom message. Pi maps it to the same user-role model input; role-changing transitions retain normal prompt preparation.\n\nCo-Authored-By: PI/OpenAI <288921227+claudypoo@users.noreply.github.com>
Require explicit human takeover confirmation before releasing an inherited or stopped worker binding. Preserve saved history and reviews, then open the replacement through the correlated worker path.\n\nCo-Authored-By: PI/OpenAI <288921227+claudypoo@users.noreply.github.com>
Require interactive workers to use the correlated OpenGoalWorker path, reject mismatched attachments, and queue stop supervision even while the parent is busy. Preserve direct human control of worker conversations and models.\n\nCo-Authored-By: PI/OpenAI <288921227+claudypoo@users.noreply.github.com>
