[Jammy2211]

Summary

Closes #341 (batch phase 6, epic two-slot-batching).

Both dashboards now open with a batch status box: the open slot, every member on one line as in progress / awaiting review / ruled, and a Review this batch → button to the Pages packet once there is something to review. One stdlib-only module, agents/conductors/batch/_status.py, computes and renders it for both organs, so the Cortex's bare Pages job (no pip install) and the Mind's dashboard stay byte-consistent, and nothing in the box is time-derived (the --check runs stay deterministic).

• Mind (dev) — behaviour unchanged. Dispatched at once, reviewed at once; the button appears when collected: is stamped; the slot closes on review.
• Cortex (science) — rolling. The button appears as soon as one member is awaiting-ruling; submitted/running members show as in progress with the record's own wall/budget/last-refresh note and hold no review control. A slot stays open across partial reviews: later sittings land as batches/reviews/<slot>-r<N>.md, batch collect --kind cortex merges them (ruled sticky), keeps refreshing the packet while anything is on the board, never re-splices a ruled member, and closes the slot only when nothing is left in a board state (members named on a - carried: line excluded). The dev refusal on an existing review file stays.

What changed

• agents/conductors/batch/_status.py (new): vocabulary, lossless read_record, dev_status/cortex_status, pick_slot by latest dispatched: (never lexical — -night sorts before -pm), render_html/render_md, fixture "No batch in flight."
• agents/conductors/batch/_batch.py: imports the vocabulary back (one definition); _review_paths/_next_review_path/_merge_reviews; organ_for offers the next free review path (packet.js unchanged — it reads REVIEW_PATH); record_update_cortex closes on the board and repeats - review: per sitting; _apply_packet refuses only for dev.
• agents/conductors/cortex/_cortex.py: census batch reading joins each record's members to the phase files' live State:; box rendered above "awaiting" in both twins; local read_record replaced by the shared one.
• agents/conductors/intake/_intake.py: _batch_status(mind); box rendered above "Start here" in both twins.
• board/_theme.py: a.go button rule.
• agents/conductors/batch/AGENTS.md: the partial-review rule, -r<N> naming, board-based close; dev rule restated as unchanged.

Companion PRs — merge order

1. cortex: accept numbered partial review files for a rolling slot (#341) PyAutoCortex#7 — cortex.py check accepts -r<N> (must land before this).
2. This PR.
3. PyAutoMind — dashboard_refresh.yml triggers on batches/** + one doctrine sentence (opened alongside; link in the issue).

Both organs' committed dashboard.md predate the box; their dashboard_refresh.yml self-heal will regenerate them after merge.

Test plan

• python -m pytest tests -q in PyAutoBrain — 861 passed (new test_batch_status.py; rolling-slot cases in test_batch_kinds.py; box tests in test_cortex_conductor.py, test_intake_dashboard.py)
• Real Cortex and Mind dashboard renders on the worktree show the fixture (nothing open) and are byte-idempotent across two runs
• After merge: both dashboards regenerate with the box; next Cortex slot with a mixed running/awaiting board shows the button with running members holding no control

🤖 Generated with Claude Code

https://claude.ai/code/session_01SgtS7NhqeziqgxKfSiMzRL

[Jammy2211]

CI red is the merge-order dependency, not a regression. The one failing test, test_batch_kinds.py::test_a_second_partial_review_refreshes_the_packet_untouched_elsewhere, runs cortex.py check from the workflow's PyAutoCortex checkout, which is main and still rejects reviews/<slot>-r2.md, so the record never gains its second - review: line. Reproduced locally: the same test fails with PYAUTO_CORTEX pointed at Cortex main and passes against the PyAutoLabs/PyAutoCortex#7 branch (full suite 861 passed there).

Merge PyAutoLabs/PyAutoCortex#7 first, then re-run this PR's checks.