feat(cortex): 📋 retire chip on dormant rows, backed by a cortex.py retire door #358

Affected Repositories

• PyAutoBrain (primary) — agents/conductors/cortex/_cortex.py, tests/test_cortex_conductor.py, skills/cortex/cortex.md
• PyAutoCortex — scripts/cortex.py, tests/test_cortex.py, REFERENCE.md, AGENTS.md, projects.yaml header comment, regenerated dashboard.md / dashboard.html

Branch Survey

Suggested branch: feature/cortex-retire-project-chip

Design decisions

• Retiring keeps the row: projects.yaml is the only record of where a project's data lives.
• Guard is "nothing live": states outside RULED_STATES ∪ {planned} block it; planned phases are unasked questions and may stay.
• Chip on dormant rows only; a planned project with nothing open has not started and is not offered retirement.

Implementation Steps — PyAutoCortex

1. PROJECT_STATUSES = ("active", "dormant", "planned", "retired"); _finish_row message lists four.
2. retire_project(root, key, why, today) beside move_phase: unknown / already-retired → CortexError; scan the project's phases with the loader check uses and refuse naming every rel — state outside RULED_STATES ∪ {"planned"}; line-edit the <key>: block — ^ status: .*$ → status: retired, note: → "retired <today>: <why>" (rewrite or append; quoted, no escapes); re-parse and require problems == [], restoring the original text on drift.
3. cmd_retire + sub.add_parser("retire", …) with project, --why (required), _common(dated=True).
4. Docs: REFERENCE.md vocabulary + "Retiring a project" paragraph; AGENTS.md verb list if it enumerates commands; projects.yaml header comment.
5. Tests beside test_planned_is_a_legal_status: retired legal; flip + note with every other byte identical; refuses on a live phase naming it; allows planned + accepted; refuses unknown / already-retired; check clean after; CLI round-trip.

Implementation Steps — PyAutoBrain

6. STATUS_RANK["retired"] = 3.
7. _retire_payload(key) beside _checkin_payload: the /cortex paste — confirm nothing live, run python3 scripts/cortex.py retire <key> --why "<one line on why>", check, dashboard --apply, push the ledger.
8. Nothing-open table in both twins: split folded into retired vs rest. Fourth column Retire: HTML <td><button class="copy" data-cmd="…" aria-label="Copy the retire prompt">📋 retire</button></td> on dormant rows, empty cell otherwise; markdown cell retire ↓ with one _task_row(f"retire {key}", payload) per dormant project under the table (a <details> inside a table cell does not render on GitHub). Retired projects: one folded <details><summary>N retired</summary>…</details> after the table carrying each row's note.
9. census --by-project "Dormant, nothing open:" printout gets a separate "Retired:" line.
10. skills/cortex/cortex.md: appendix row for retire <project> --why "…" and a rule bullet — retiring never deletes a row or a ruling.

Tests — PyAutoBrain

• Fixture's dormant nothing-open project carries the chip (HTML data-cmd contains cortex.py retire <key> --why; markdown has the 📋 retire <key> row); a planned row has an empty cell; a retired row is absent from the table, present in the "1 retired" fold with its note, ordered last.
• Run with PYAUTO_CORTEX=<worktree>/PyAutoCortex; full Brain suite once; Cortex suite plain.

Key Files

• PyAutoCortex/scripts/cortex.py — PROJECT_STATUSES, _finish_row, move_phase (pattern), parse_projects / load_projects, parser
• PyAutoBrain/agents/conductors/cortex/_cortex.py — STATUS_RANK, by_project_keys, _checkin_payload, _html_task, _task_row, Nothing-open table in both twins
• PyAutoBrain/skills/cortex/cortex.md — verb appendix

Next to dormant projects can you put a copy thing which I would put into claude to retire it

Prompt file: PyAutoMind/draft/feature/pyautobrain/cortex_retire_project_chip.md