Tasks: Restore green main: stale test premises

Mission: main-red-test-premise-repair-01M4XV67 · Topology: single_branch (WPs run sequentially in the write checkout)

Subtask Index

IDDescriptionWPParallel
T001Record red-first proof for both failing tests at baseWP01
T002Re-pin test_is_well_formed_declaration over (token, mission_type, expected)WP01
T003Add research-kind and placeholder casesWP01
T004Apply the sanctioned corrections to build_recovery_repo(full=True) as a second fixture commitWP01
T005Run the recovery, byte-identity and lint targeted suitesWP01
T006Add the uv.lock-only non-src filter group and output to ci-routerWP02[P]
T007Fan the group out to upgrade and status through a separate non-census DEPENDENCY_MODULE_FANOUT mapWP02
T008Lockstep pins: workflow coherence, gate-selection authority, fan-out census/allowlistWP02
T009Focused selection test: lock-only selects upgrade+status; manifest-only adds no upgradeWP02

WP01 – Re-pin the two stale test premises (#6240, #6238)

Prompt: tasks/WP01-repin-stale-test-premises.md · Priority: P1 · Dependencies: none

Goal: both red tests pass on the pinned events version, with the audit as strict as before.

T001 Record red-first proof (WP01) T002 Re-pin the well-formedness test signature (WP01) T003 Research-kind and placeholder cases (WP01) T004 Sanctioned-correction overlay in the full-corpus fixture (WP01) T005 Targeted test runs (WP01)

Risk: the overlay could perturb other assertions that compare against the original bytes. Mitigation: run the whole recovery and byte-identity files.

WP02 – Lock-file reducer bump selects upgrade and status shards

Prompt: tasks/WP02-lockfile-fanout-routing.md · Priority: P2 · Dependencies: none

Goal: a uv.lock change selects the upgrade and status module shards; pyproject.toml is excluded (operator ruling).

T006 ci-router filter group + output (WP02) T007 DEPENDENCY_MODULE_FANOUT map (non-census) (WP02) T008 Lockstep pins (WP02) T009 Focused selection test (WP02)

Risk: the transcription and census guards red if any lockstep site is missed. Mitigation: a separate non-census map keeps the pack census untouched; a mutation proof shows the test is not vacuous.