Affected Repositories

• PyAutoMemory (primary)

Branch Survey

Suggested branch: feature/memory-state-feed
Worktree: ~/Code/PyAutoLabs-wt/organ-cockpit-feeds/ (shared with the Hands member)

Implementation Steps

1. scripts/board.py — to_state(snapshot) -> dict (stdlib): status grey when _totals(snapshot)["pages"] == 0; yellow if the inbox or interests digest is stale (_inbox_freshness(...)["stale"]) or any wiki carries stubs/todos; else green. Memory has no red condition (nothing can break) — the docstring says so. headline = the badge message (N pages · P% cited). updated = snapshot generated as whole-second …Z, unparseable → now. pages_url = pages_url(snapshot). items = stale inbox/interests → yellow with the board URL; each wiki with stubs/todos → info with the existing _stub_prompt/_todo_prompt payload; inbox papers waiting → info with _queue_issue_url. fmt == "state" in render; --state in the mutually-exclusive group with --badge.
2. .github/workflows/knowledge_board.yml — render step adds python scripts/board.py --state > _site/state.json; new step "Validate the cockpit feed against the Brain contract": python3 PyAutoBrain/board/_state.py _site/state.json (PyAutoBrain is already checked out at ./PyAutoBrain). Header comment lists state.json beside badge.json.
3. tests/test_board.py — on the synthetic tree fixture: required keys, status enum, updated ends with Z and parses, stubs → info items with prompts, stale inbox → yellow, empty tree → grey, render(fmt="state") renders. Copy the 6-line required/enum check; do not import Brain. Tenant firewall with --root <worktree> must be OK.

Key Files

• scripts/board.py — renderer (badge_endpoint, _totals, _inbox_freshness, _queue_issue_url, _stub_prompt, _todo_prompt, pages_url)
• .github/workflows/knowledge_board.yml — render + validate steps
• tests/test_board.py

Verification

• pytest tests/test_board.py -q; python scripts/board.py --state | python ../PyAutoBrain/board/_state.py - → state: ok
• After merge: knowledge_board.yml run green with the validate step logging state: ok; curl https://pyautolabs.github.io/PyAutoMemory/state.json validates.

Organ cockpit: PyAutoMemory knowledge board emits state.json

Type: feature
Target: PyAutoMemory
Repos:

• PyAutoMemory
  Difficulty: small
  Autonomy: safe
  Priority: high
  Status: formalised
  Consequence: notify
  Witness: knowledge_board.yml run on main green with the validate step logging 'state: ok'; curl https://pyautolabs.github.io/PyAutoMemory/state.json passes the validator CLI.
  Review-minutes: 0
  Unattended: ready
  Epic: organ-cockpit

Phase 1 (issue 416 in the Brain, shipped 2026-09-26) defined the per-organ state.json v1 cockpit feed (contract board/state_schema.json, validator + CLI board/_state.py in the Brain checkout the board job already has at ./PyAutoBrain). This prompt makes the Memory knowledge board emit it, mirroring the Heart pattern.

1. scripts/board.py in PyAutoMemory: a --state flag beside --badge rendering the v1 feed: status from the badge colour (red/orange/lightgrey/brightgreen -> red/yellow/grey/green), headline = the badge message, updated = the render timestamp as ISO-8601 Z, pages_url = https://pyautolabs.github.io/PyAutoMemory/, items = the actionable rows the board already renders (stale wiki pages, unprovenanced entries, reading-queue waits) with GitHub links.
2. .github/workflows/knowledge_board.yml: python scripts/board.py --state > _site/state.json in the render step, then a validate step python3 PyAutoBrain/board/_state.py _site/state.json that fails the job on drift.
3. Tests beside the existing board tests in PyAutoMemory: required keys, status enum, updated ends with Z; Memory tests must not import the Brain.

Witness: knowledge_board.yml run on main green with the validate step logging 'state: ok'; curl https://pyautolabs.github.io/PyAutoMemory/state.json passes the validator CLI.