A cell/main PM's i_will_plan could legally re-claim its own task from
awaiting_pm_review (a CLAIM_RULES edge added for post-respawn recovery),
resetting the task to in_progress and re-running submit_up -> pr_pass ->
awaiting_pm_review forever: one Sentinel conventions child looped eleven
full laps in four hours (14 reviews on one PR, 37 agent spawns) while
its root's closure check fired eighteen times and the Main PM could
never close anything.
The claim edge is gone from every table that carried it — CLAIM_RULES,
the claim ActionSpec's source statuses, the StatusTransition row, the
service-layer _ROLE_CLAIM_STATUSES twin, and the legacy enforcement
shim's operational-edge/role-gate entries (left divergent, it would be
the same silent two-table drift that produced this bug). The respawn
case the edge existed for is now served properly: _handle_pm_reentry
gained a third contract — a PM calling i_will_plan on its own
awaiting_pm_review task gets a steering envelope (no claim, no state
change) pointing at complete/request_changes, and give_me_work's next
hint for that status says the same instead of steering back into
i_will_plan. The no-transition review-claim path and the pm-review
dispatch prompt were already correct and are untouched, so closure
still converges through them.
A new bidirectional test asserts CLAIM_RULES and _ROLE_CLAIM_STATUSES
stay identical per PM role (the old comment claimed a sync test existed;
it checked one direction only). Lifecycle artifacts regenerated; the
parity suite's three unshaped session mocks fixed, zero AsyncMock
warnings remain.
make quality runs 'mypy roboco/ tests/'; the new PM-needs_revision claim test passed a SimpleNamespace where _get_valid_claim_statuses expects AgentTable | None, failing the type gate. The helper only reads agent.role, so cast the lightweight stand-in to AgentTable (no DB row, no type: ignore).
The lifecycle spec (CLAIM_RULES) grants CELL_PM/MAIN_PM claim of NEEDS_REVISION so a rejected coordination root (pr_fail / qa_fail / ceo_reject) can be re-claimed via i_will_plan and re-delegated. The runtime mapping _ROLE_CLAIM_STATUSES omitted it for PMs, so the spec gate allowed i_will_plan on a needs_revision root while the composed claim() rejected it -> returned None -> INVALID_STATE: the PM could neither plan nor idle its own rejected root and respawn-looped (observed live on cell root 0e49e04e, ~143 INVALID_STATE rejections across 11 PM sessions; the tail of the 2026-06-24/25 run).
Add NEEDS_REVISION to the cell_pm/main_pm runtime claim statuses, and a parity test locking _ROLE_CLAIM_STATUSES to lifecycle.CLAIM_RULES so the two can't drift again.
Renn F