Tasks: Test-Suite Friction Remediation

Mission: test-suite-friction-remediation-01KXDKBX | Branch: feat/test-suite-friction-remediation Plan: plan.md (14-IC / 4-lane map, post-plan-squad hardened) | Spec: spec.md

17 work packages across 4 lanes. Every WP is test-first with a non-fakeable definition of done (NFR-002) and every WP that touches or observes a ratchet/pinning or behavioural-parity suite appends a row to tracer-design-decisions.md (the standing catalog) + logs friction to tracer-tooling-friction.md (FR-016 / NFR-007).

Lane shape (binding — see plan.md "Lane & WP shaping directives")

(delete) → WP05. All co-tenant the dead-code gate's allowlist surface; they MUST run as one ordered chain, never concurrently. NFR-001: the dead-code gate (WP01) learns first-party dynamic access before any Cluster-0 deletion (WP18/WP05). Re-sequence note: the runtime_bridge src-delegate deletion was moved OUT of WP02 into the new WP18 (post-repoint), because a delegate cannot be deleted until WP03/WP04 repoint the test sites and the frozen compat baseline is updated — WP02 empirically proved 0-deletable-today.

(directory-disjoint conversion batches, each deps WP11).

baseline burden) → WP16 (shard-registry seam) → WP17 (CI gate + coverage guards).

  • Lane 0 — deshim/tooling (STRICT SERIAL CHAIN): WP01 → WP02 (classify-only) → WP03 → WP04 → WP18
  • Lane A — test-intrinsic (PARALLEL roots): WP06, WP07, WP08, WP09, WP10 — independent, no cross edges.
  • Lane C — golden-count sweep: WP11 (inventory + recurrence guard, deps WP07) → WP12, WP13, WP14
  • Lane B — CI guards (SERIAL): WP15 (gc2b scope-to-orphan, root — lands early to relieve the new-file

Single-file ownership invariants: tests/architectural/test_no_dead_symbols.py → WP05 only; tests/conftest.py → WP16 only; .github/workflows/ci-quality.yml → WP17 only; tests/_arch_shard_map.py → WP16 owns the seam (WP11/WP17 append their new-guard registration as recorded out-of-map edits); src/runtime/next/runtime_bridge.py + tests/runtime/test_bridge_compat_surface.py → WP18 owns the deletion (WP04's tests/runtime/* glob overlaps the compat file but is sequential-exempt: WP18 ← WP04).

Dependency graph (edges)

WP01 → WP02 → WP03 → WP04 → WP18 → WP05  (Lane 0 serial; WP02 classify-only, WP18 delete)
WP06  (root)   WP07 (root)   WP08 (root)   WP09 (root)   WP10 (root)   (Lane A)
WP07 → WP11 → {WP12, WP13, WP14}        (Lane C)
WP15 → WP16 → WP17                       (Lane B serial)

Roots: WP01, WP06, WP07, WP08, WP09, WP10, WP15.

Subtask Index

IDDescriptionWPParallel
T001First-party module.attr dynamic-access resolution in _symbol_key.py (general accessor shape)WP01
T002Fixture-based AST test: dynamic-access → live (into test_no_dead_symbols.py, recorded out-of-map)WP01
T003Fixture-based AST test: unreferenced → dead (negative direction still holds)WP01
T004Verify the 4 known façade symbols classify live without an allowlist row; gate greenWP01
T005ruff/mypy clean; tracer catalog + friction rowsWP01
T006Grep-classify the ~416 setattr(runtime_bridge…) occurrences (53 files; WP03 ~241 + WP04 ~168) forwarding-vs-real-seamWP02
T007Delete the pure re-export delegates in runtime_bridge.py; keep the 8 canonical __all__ namesWP02
T008Delete pure re-export delegates in the seam modules (if any)WP02
T009Dead-code gate green (proves each deleted delegate dead) + runtime suite greenWP02
T010ruff/mypy clean; NFR-001 dependency-on-WP01 recorded; tracer rowsWP02
T011Repoint forwarding monkeypatch sites in tests/next/** to their owning seam modulesWP03
T012Repoint the two heavy files (test_query_mode_unit, test_runtime_bridge_unit)WP03
T013tests/next suite green; no runtime_bridge-forwarded patch remainsWP03
T014ruff/mypy clean; tracer rowsWP03
T015Repoint forwarding sites in tests/runtime/, tests/specify_cli/WP04
T016Repoint forwarding sites in tests/integration/, tests/contract/, miscWP04
T017Affected suites green (incl. tests/runtime/test_bridge_parity.py)WP04
T018ruff/mypy clean; tracer rows (parity suite observed)WP04
T019Remove the 4 runtime_bridge façade allowlist rows (now live via WP01 gate)WP05
T020Burn down the category_b_grandfathered_legacy subset WP01 reclassifies as genuinely deadWP05
T021Single baseline recount (193 → decreases, NOT to 0); keep live doctrine.* re-exportsWP05
T022DoD: dead-code gate green; git grep <deleted symbol> = 0 across src+testsWP05
T023ruff/mypy clean; tracer catalog rowsWP05
T024Extend the ban guard to catch (rel,int) laundered via loop var into composite_key(source,N)WP06[P]
T025Regression fixture that attempts the laundering fails the banWP06[P]
T026Convert test_trio_seam_only._IO_ALLOWLIST_SITES (~L463) to content-addressed keysWP06[P]
T027DoD: git grep -E '\.py", *[0-9]{3}\)' tests/architectural/ = 0; ban green on the real treeWP06[P]
T028ruff/mypy clean; tracer rowsWP06[P]
T029Assert the exact Lane member-name frozenset (replace len(Lane) == N)WP07[P]
T030Rename the test off "…nine_values"WP07[P]
T031tests/status/test_models.py greenWP07[P]
T032ruff/mypy clean; tracer row (exemplar pattern for WP11)WP07[P]
T033Re-point the confirmed twin (:211 read_text) to an observable contract, no @patch on the SUTWP08[P]
T034Audit test_dashboard/test_api_handler.py — repoint ONLY if a genuine source-as-text twinWP08[P]
T035Audit glossary/test_event_emission.py + sync/tracker/test_service.py — repoint only genuine twinsWP08[P]
T036DoD: no @patch on the SUT in any re-pointed test; assert a persisted artifactWP08[P]
T037ruff/mypy clean; tracer parity rowsWP08[P]
T038Add a side-effect-free / no-coordination entrypoint to create_mission_core() if missingWP09[P]
T039Implement make_mission() delegating wrapper in tests/_factories/__init__.pyWP09[P]
T040Factory-parity test: meta.json byte-identical (minus overrides) to a direct core callWP09[P]
T041DoD: tests/_factories/__init__.py non-empty with ≥1 real importerWP09[P]
T042ruff/mypy clean; tracer row (parity assertion)WP09[P]
T043Verify #2553 is a real contract fix, not a warning suppression → close or minimal-fixWP10[P]
T044Recount #2295 quarantine (~1 active marker, test_summary_tolerance.py:704)WP10[P]
T045Triage the residual marker; cross-ref #2309; do NOT drive to 0; MUST NOT edit conftest.pyWP10[P]
T046DoD evidence; confirm zero production behaviour changeWP10[P]
T047tracer row (quarantine ratchet observed)WP10[P]
T048AST inventory of len(<collection>) == <int> in tests/ → classify keep vs convertWP11
T049Write the recurrence guard test_golden_count_ban.py (fails NEW un-annotated; escape hatch)WP11
T050Regression: a fresh un-annotated len==int fixture fails the guardWP11
T051Establish the convert-set baseline; emit the inventory artifact partitioned BY DIRECTORYWP11
T052New-guard-file DoD: register in _arch_shard_map.py (out-of-map append) + refreeze gc3b & gc2bWP11
T053DoD: guard green on the real tree; anchors WP07's exemplar patternWP11
T054ruff/mypy clean; tracer catalog rowsWP11
T055Pull WP11's inventory convert-set for this batch's owned directoriesWP12[P]
T056Convert len==N → set/frozenset-equality per file (owned dirs only)WP12[P]
T057Decrement the convert-set baseline (strictly decreases; never regrows)WP12[P]
T058Affected suites + golden-count guard greenWP12[P]
T059ruff/mypy clean; tracer rowsWP12[P]
T060Pull WP11's inventory convert-set for this batch's owned directoriesWP13[P]
T061Convert len==N → set/frozenset-equality per file (owned dirs only)WP13[P]
T062Decrement the convert-set baseline (strictly decreases; never regrows)WP13[P]
T063Affected suites + golden-count guard greenWP13[P]
T064ruff/mypy clean; tracer rowsWP13[P]
T065Pull WP11's inventory convert-set for this batch's owned directoriesWP14[P]
T066Convert len==N → set/frozenset-equality per file (owned dirs only)WP14[P]
T067Decrement the convert-set baseline (strictly decreases; never regrows)WP14[P]
T068Affected suites + golden-count guard greenWP14[P]
T069ruff/mypy clean; tracer rowsWP14[P]
T070Scope the gc2b exact-selection ratchet to orphans (or advisory) per #2616WP15
T071Regression: routine test-file add/remove no longer trips red; orphan detection preservedWP15
T072Verify the load-bearing orphan-detection signal is retainedWP15
T073DoD; ruff/mypy cleanWP15
T074tracer row (gc2b ratchet)WP15
T075Introduce _shard_registry.py: idempotent register() / all_groups() + expected-group manifestWP16
T076Refactor _arch_shard_map.py + _next_shard_map.py to register via the seam (drop # noqa: F401)WP16
T077test_arch_shard_marker_completeness.py: diagnosable "group not registered", never bare KeyErrorWP16
T078conftest.py shard-marker hook uses the registry; an unmarked tests/next universe fails loudWP16
T079Regression: drop _next_shard_map registration → diagnosable message; unmarked universe failsWP16
T080New-guard-file DoD (self-register + refreeze baselines); ruff/mypy; tracerWP16
T081test_suite_jobs_gate_blocking.py: pytest_jobs − NON_BLOCKING_ALLOWLIST ⊆ quality-gate.needsWP17
T082Resolve slow-tests / mutation-testing to explicit gated OR allowlisted-with-rationaleWP17
T083Regression: a fake pytest job added to the model → guard failsWP17
T084ci-quality.yml sonarcloud discovers ui-e2e.yml coverage-ui-e2e.xml (cross-workflow, head-SHA)WP17
T085test_ui_e2e_coverage_discovered.py wiring guard asserts the path is in the discovered setWP17
T086ci-quality.yml quality-gate.needs edits if a job must join; new-guard-file DoD; ruff/mypy; tracerWP17
T087Delete the WP02-classified forwarding delegates now repointed by WP03/WP04 (grep=0 per name; gate proves dead)WP18
T088Refactor the 3 production _rb.<name> call-sites to import the seam leaf directly; delete their façade delegatesWP18
T089Update the frozen test_bridge_compat_surface.py exact-baseline: drop exactly the deleted symbols; surviving sentinels intactWP18
T090Dead-code gate + tests/runtime (incl. test_bridge_parity.py) greenWP18
T091ruff/mypy clean; final runtime_bridge parity/compat tracer verdict; NFR-001 recordedWP18

Work Packages

WP01 — Dead-code gate: first-party dynamic-access awareness (FR-001)

Goal: Teach _symbol_key.py to resolve first-party module.attr dynamic access as a live reference, generally (no runtime_bridge special-case). Pure tooling — no allowlist edits (those are WP05's). NFR-001 anchor: this lands before any Cluster-0 deletion. Independent test: focused AST fixtures prove both directions (dynamic-access→live, unreferenced→dead).

  • ✅ T001 First-party module.attr dynamic-access resolution in _symbol_key.py (WP01)
  • ✅ T002 Fixture AST test: dynamic-access → live (into test_no_dead_symbols.py, recorded out-of-map) (WP01)
  • ✅ T003 Fixture AST test: unreferenced → dead (WP01)
  • ✅ T004 The 4 known façade symbols classify live without a row; gate green (WP01)
  • ✅ T005 ruff/mypy clean; tracer catalog + friction rows (WP01)

Prompt: tasks/WP01-dead-code-dynamic-access.md · ~240 lines · deps: none

WP02 — Classify the runtime_bridge compat-delegate surface (FR-003) — CRITICAL PATH

Goal (re-sequenced → classification-only): Grep-classify the ~416 setattr(runtime_bridge…) occurrences (53 test files; WP03 ~241 in tests/next + WP04 ~168) into forwarding-vs-real-seam, and record the 45-candidate pure-forwarding table (deletable-after-repoint vs canonical-__all__ vs needs-call-site-refactor). The src-delegate DELETION moved to WP18 (post-repoint) — WP02 empirically proved 0-deletable-today (deleting any forwarder reds the frozen test_bridge_compat_surface.py baseline and/or an un-repointed patcher; 3 have live production _rb.<name> deps). Independent test: classification recorded + baseline stays green (zero src diff).

  • ✅ T006 Grep-classify the ~416 forwarding-vs-real-seam occurrences (53 files; WP03 ~241 + WP04 ~168) (WP02)
  • ✅ T007 Delete pure re-export delegates in runtime_bridge.py; keep the 8 __all__ names (WP02)
  • ✅ T008 Delete pure re-export delegates in the seam modules (if any) (WP02)
  • ✅ T009 Dead-code gate green + runtime suite green (WP02)
  • ✅ T010 ruff/mypy clean; NFR-001 dep recorded; tracer rows (WP02)

Prompt: tasks/WP02-runtime-bridge-delegate-deletion.md · ~260 lines · deps: WP01

WP03 — Repoint monkeypatch sites, batch A: tests/next/** (FR-003)

Goal (repoint-only): Repoint the forwarding monkeypatch.setattr(runtime_bridge, …) sites in tests/next/ (12 files, ~241 occurrences incl. the two heaviest) at their owning seam modules, so the delegates become deletable by WP18. Do NOT delete any src delegate here (that is WP18's surface, post-WP04). Independent test**: tests/next suite green with zero remaining runtime_bridge-forwarded patches.

  • ✅ T011 Repoint forwarding sites in tests/next/** (WP03)
  • ✅ T012 Repoint the two heavy files (test_query_mode_unit, test_runtime_bridge_unit) (WP03)
  • ✅ T013 tests/next suite green (WP03)
  • ✅ T014 ruff/mypy clean; tracer rows (WP03)

Prompt: tasks/WP03-repoint-monkeypatch-batch-a.md · ~220 lines · deps: WP02

WP04 — Repoint monkeypatch sites, batch B: runtime/specify_cli/integration/misc (FR-003)

Goal (repoint-only): Repoint the remaining forwarding sites (runtime, specify_cli, integration, contract, agent, perf, unit — ~40 files, ~168 occurrences), disjoint from WP03, so the delegates become deletable by WP18. Do NOT delete any src delegate here, and do NOT edit tests/runtime/test_bridge_compat_surface.py (WP18 owns the frozen-baseline update; WP04's tests/runtime/ glob is sequential-exempt from it). Independent test**: affected suites green incl. the behavioural-parity tests/runtime/test_bridge_parity.py.

  • ✅ T015 Repoint tests/runtime/, tests/specify_cli/ (WP04)
  • ✅ T016 Repoint tests/integration/, tests/contract/, misc (WP04)
  • ✅ T017 Affected suites green (incl. test_bridge_parity.py) (WP04)
  • ✅ T018 ruff/mypy clean; tracer rows (parity suite observed) (WP04)

Prompt: tasks/WP04-repoint-monkeypatch-batch-b.md · ~230 lines · deps: WP03

WP05 — Allowlist reconciliation + grandfathered burndown (FR-002, FR-004)

Goal: Remove the 4 façade allowlist rows (now live via WP01) and burn down the category_b_grandfathered_legacy subset WP01 reclassifies as genuinely dead — baseline 193 decreases, NOT to 0; keep live doctrine.* re-exports. Independent test: dead-code gate green; git grep <deleted symbol> = 0 across src+tests.

  • ✅ T019 Remove the 4 runtime_bridge façade allowlist rows (WP05)
  • ✅ T020 Burn down the category_b_grandfathered_legacy genuinely-dead subset (WP05)
  • ✅ T021 Single baseline recount (193 → decreases, not 0); keep live doctrine.* (WP05)
  • ✅ T022 DoD: gate green; git grep <deleted symbol> = 0 across src+tests (WP05)
  • ✅ T023 ruff/mypy clean; tracer catalog rows (WP05)

Prompt: tasks/WP05-allowlist-reconciliation-burndown.md · ~240 lines · deps: WP18

WP06 — Close the seed-tuple laundering hole in the positional-anchor ban (FR-005)

Goal: Extend test_ratchet_positional_anchor_ban.py to flag a (rel,int) seed laundered through a loop var into composite_key(source,N), and convert test_trio_seam_only._IO_ALLOWLIST_SITES (~L463) to content-addressed keys. test_no_write_side_rederivation.py is already clean — do NOT touch it. Independent test: git grep -E '\.py", *[0-9]{3}\)' tests/architectural/ = 0; ban green on the real tree.

  • ✅ T024 Extend the ban guard to catch the loop-var laundering (WP06)
  • ✅ T025 Regression fixture attempting the laundering fails the ban (WP06)
  • ✅ T026 Convert test_trio_seam_only._IO_ALLOWLIST_SITES (~L463) to content-addressed keys (WP06)
  • ✅ T027 DoD: git grep for positional anchors = 0; ban green on the real tree (WP06)
  • ✅ T028 ruff/mypy clean; tracer rows (WP06)

Prompt: tasks/WP06-positional-anchor-ban-hole.md · ~230 lines · deps: none

WP07 — Lane-enum golden count → exact frozenset (FR-006)

Goal: Replace len(Lane) == N with an exact Lane member-name frozenset and rename the test off "…nine_values". The exemplar pattern for WP11's sweep. Independent test: adding/removing a lane forces a content edit; tests/status/test_models.py green.

  • ✅ T029 Assert the exact Lane member-name frozenset (WP07)
  • ✅ T030 Rename the test off "…nine_values" (WP07)
  • ✅ T031 tests/status/test_models.py green (WP07)
  • ✅ T032 ruff/mypy clean; tracer row (exemplar for WP11) (WP07)

Prompt: tasks/WP07-lane-enum-frozenset.md · ~200 lines · deps: none

WP08 — Source-as-text wiring → observable contract (+ sibling audit) (FR-007)

Goal: Re-point the confirmed read_text() twin (:211) to an observable contract with no @patch on the SUT; audit the 3 siblings and re-point ONLY genuine twins (do not manufacture a false shared helper). Independent test: no @patch on the SUT in any re-pointed test; assertion on a persisted artifact.

  • ✅ T033 Re-point the confirmed twin (:211 read_text) to an observable contract (WP08)
  • ✅ T034 Audit test_dashboard/test_api_handler.py; repoint only if a genuine twin (WP08)
  • ✅ T035 Audit glossary/test_event_emission.py + sync/tracker/test_service.py; repoint only genuine twins (WP08)
  • ✅ T036 DoD: no @patch on the SUT; assert a persisted artifact (WP08)
  • ✅ T037 ruff/mypy clean; tracer parity rows (WP08)

Prompt: tasks/WP08-source-as-text-observable.md · ~240 lines · deps: none

WP09 — Production-delegating mission factory (FR-008)

Goal: tests/_factories.make_mission() delegates to create_mission_core(); add a side-effect-free / no-coordination entrypoint to the core if missing (behaviour-preserving). Independent test: output meta.json byte-identical (minus overrides) to a direct core call; _factories/__init__.py non-empty with ≥1 real importer.

  • ✅ T038 Add a side-effect-free / no-coord entrypoint to create_mission_core() if missing (WP09)
  • ✅ T039 make_mission() delegating wrapper in tests/_factories/__init__.py (WP09)
  • ✅ T040 Factory-parity test: meta.json byte-identical minus overrides (WP09)
  • ✅ T041 DoD: _factories/__init__.py non-empty with ≥1 real importer (WP09)
  • ✅ T042 ruff/mypy clean; tracer row (WP09)

Prompt: tasks/WP09-mission-factory-delegation.md · ~230 lines · deps: none

WP10 — Verify-and-close hygiene: #2553 + #2295 (FR-009, FR-010)

Goal: Verify #2553 is a real fix (not a suppression) → close or minimal-fix; recount #2295 quarantine (~1 active marker), triage it, cross-ref #2309, do NOT drive to 0. MUST NOT edit tests/conftest.py. Independent test: DoD evidence recorded; zero production behaviour change.

  • ✅ T043 Verify #2553 is a real fix → close or minimal-fix (WP10)
  • ✅ T044 Recount #2295 quarantine (~1 marker, test_summary_tolerance.py:704) (WP10)
  • ✅ T045 Triage the residual marker; cross-ref #2309; no drive-to-0; no conftest.py edit (WP10)
  • ✅ T046 DoD evidence; confirm zero production change (WP10)
  • ✅ T047 tracer row (quarantine ratchet observed) (WP10)

Prompt: tasks/WP10-verify-close-hygiene.md · ~220 lines · deps: none

WP11 — Golden-count inventory + recurrence guard (FR-014)

Goal: AST inventory of len(<collection>) == <int> in tests/ (classify keep vs convert), write the recurrence guard test_golden_count_ban.py (fails on NEW un-annotated golden-count; escape hatch # golden-count: cardinality-is-contract), and emit the convert-set inventory partitioned BY DIRECTORY for the batch WPs. New-guard-file DoD applies. Independent test: a fresh un-annotated len==int fixture fails the guard; guard green on the real tree.

  • ✅ T048 AST inventory of len(...) == int in tests/ → classify keep vs convert (WP11)
  • ✅ T049 Recurrence guard test_golden_count_ban.py (fails NEW un-annotated; escape hatch) (WP11)
  • ✅ T050 Regression: a fresh un-annotated fixture fails the guard (WP11)
  • ✅ T051 Convert-set baseline + inventory artifact partitioned BY DIRECTORY under feature_dir (WP11)
  • ✅ T052 New-guard-file DoD: register in _arch_shard_map.py (out-of-map append) + refreeze gc3b & gc2b (WP11)
  • ✅ T053 DoD: guard green on the real tree; anchors WP07's exemplar pattern (WP11)
  • ✅ T054 ruff/mypy clean; tracer catalog rows (WP11)

Prompt: tasks/WP11-golden-count-inventory-guard.md · ~290 lines · deps: WP07

WP12 — Golden-count conversion batch 1: doctrine & charter (FR-014)

Goal: Convert the WP11-inventory convert-set in tests/charter/, tests/doctrine/, tests/doctrine_synthesizer/, tests/glossary/. Independent test: convert-set baseline strictly decreases; owned-dir suites + golden-count guard green.

  • ✅ T055 Pull WP11's inventory convert-set for the owned directories (WP12)
  • ✅ T056 Convert len==N → set/frozenset-equality per file (owned dirs only) (WP12)
  • ✅ T057 Decrement the convert-set baseline (strictly decreases) (WP12)
  • ✅ T058 Affected suites + golden-count guard green (WP12)
  • ✅ T059 ruff/mypy clean; tracer rows (WP12)

Prompt: tasks/WP12-golden-count-batch-doctrine-charter.md · ~210 lines · deps: WP11

WP13 — Golden-count conversion batch 2: lifecycle & upgrade (FR-014)

Goal: Convert the convert-set in tests/upgrade/, tests/dossier/, tests/lanes/, tests/migration/, tests/migrate/, tests/post_merge/, tests/merge/, tests/coordination/, tests/review/. Independent test**: convert-set baseline strictly decreases; suites + guard green.

  • ✅ T060 Pull WP11's inventory convert-set for the owned directories (WP13)
  • ✅ T061 Convert len==N → set/frozenset-equality per file (owned dirs only) (WP13)
  • ✅ T062 Decrement the convert-set baseline (strictly decreases) (WP13)
  • ✅ T063 Affected suites + golden-count guard green (WP13)
  • ✅ T064 ruff/mypy clean; tracer rows (WP13)

Prompt: tasks/WP13-golden-count-batch-lifecycle-upgrade.md · ~210 lines · deps: WP11

WP14 — Golden-count conversion batch 3: cli/auth/tasks & long tail (FR-014)

Goal: Convert the convert-set in the remaining clean directories (audit, auth, tasks, missions, cross_cutting, docs, cli, doctor, core, characterization, kernel, policy, delivery, research, git_ops, event_journal, dashboard, context, ci, paths, init, e2e, cross_branch, concurrency, release, readiness, proof, mission_metadata, calibration). Independent test: convert-set baseline strictly decreases; suites + guard green.

  • ✅ T065 Pull WP11's inventory convert-set for the owned directories (WP14)
  • ✅ T066 Convert len==N → set/frozenset-equality per file (owned dirs only) (WP14)
  • ✅ T067 Decrement the convert-set baseline (strictly decreases) (WP14)
  • ✅ T068 Affected suites + golden-count guard green (WP14)
  • ✅ T069 ruff/mypy clean; tracer rows (WP14)

Prompt: tasks/WP14-golden-count-batch-cli-auth-tail.md · ~220 lines · deps: WP11

WP15 — gc2b exact-selection ratchet: scope to orphans (FR-015)

Goal: Stop the gc2b exact-selection baseline over-firing on routine test-file add/remove (this mission does it 4–5×). Scope the ratchet to orphans (or make it advisory) per #2616, preserving orphan detection. Lands early so the other new-file WPs' baseline burden shrinks. Independent test: a routine add/remove no longer trips red; the orphan-detection true-positive still fires.

  • ✅ T070 Scope the gc2b exact-selection ratchet to orphans (or advisory) per #2616 (WP15)
  • ✅ T071 Regression: routine add/remove no longer trips red; orphan detection preserved (WP15)
  • ✅ T072 Verify the load-bearing orphan-detection signal is retained (WP15)
  • ✅ T073 DoD; ruff/mypy clean (WP15)
  • ✅ T074 tracer row (gc2b ratchet) (WP15)

Prompt: tasks/WP15-gc2b-scope-to-orphan.md · ~210 lines · deps: none

WP16 — Explicit shard-registry seam (FR-011)

Goal: Replace the import-side-effect SHARD_GROUPS assembly with an idempotent register()/all_groups() + expected-group manifest so the completeness guard fails diagnosably (never a bare KeyError) and an unmarked tests/next universe fails loud. Owns tests/conftest.py. New-guard-file DoD applies. Independent test: dropping the _next_shard_map registration emits the diagnosable message; an unmarked tests/next universe fails.

  • ✅ T075 _shard_registry.py: idempotent register() / all_groups() + expected-group manifest (WP16)
  • ✅ T076 Refactor _arch_shard_map.py + _next_shard_map.py to register via the seam (WP16)
  • ✅ T077 test_arch_shard_marker_completeness.py: diagnosable "group not registered" (WP16)
  • ✅ T078 conftest.py shard-marker hook uses the registry; unmarked tests/next fails loud (WP16)
  • ✅ T079 Regression: drop _next_shard_map registration → diagnosable message; unmarked universe fails (WP16)
  • ✅ T080 New-guard-file DoD (self-register + refreeze baselines); ruff/mypy; tracer (WP16)

Prompt: tasks/WP16-shard-registry-seam.md · ~270 lines · deps: WP15

WP17 — CI gate + coverage guards (FR-012, FR-013)

Goal: Two guards over ci-quality.yml (single-file ownership): (a) every pytest-invoking job is in quality-gate.needs minus a reasoned NON_BLOCKING_ALLOWLIST; (b) the sonarcloud job discovers the ui-e2e.yml-run coverage-ui-e2e.xml (cross-workflow, head-SHA-keyed) with a wiring guard. New-guard-file DoD applies. Independent test: a fake pytest job → guard fails; the coverage path is in the discovered set.

  • ✅ T081 test_suite_jobs_gate_blocking.py: pytest_jobs − allowlist ⊆ quality-gate.needs (WP17)
  • ✅ T082 Resolve slow-tests / mutation-testing to explicit gated OR allowlisted-with-rationale (WP17)
  • ✅ T083 Regression: a fake pytest job → guard fails (WP17)
  • ✅ T084 ci-quality.yml sonarcloud discovers coverage-ui-e2e.xml cross-workflow head-SHA (WP17)
  • ✅ T085 test_ui_e2e_coverage_discovered.py wiring guard asserts the path is discovered (WP17)
  • ✅ T086 quality-gate.needs edits if a job must join; new-guard-file DoD; ruff/mypy; tracer (WP17)

Prompt: tasks/WP17-ci-gate-coverage-guards.md · ~280 lines · deps: WP16

WP18 — Retire runtime_bridge src delegates (repoint-then-delete residue) (FR-003) — LANE-0 DELETION POLE

Goal: The deletion pole re-sequenced OUT of WP02, running AFTER the WP03/WP04 repoints. Delete the WP02-classified pure-forwarding delegates in runtime_bridge.py + seam modules (canonical __all__ kept), refactor the 3 live production _rb.<name> call-sites (_retrospective_blocks_completion, _composition_dispatch_inputs, _has_generated_docs) to import the seam leaf directly, and update the FROZEN tests/runtime/test_bridge_compat_surface.py exact-baseline. Owns src/runtime/next/runtime_bridge*.py + that compat test. Independent test: dead-code gate green post-delete (each delegate proven dead) + tests/runtime (incl. test_bridge_parity.py) green; git grep <deleted symbol> = 0 across src+tests.

  • ✅ T087 Delete the WP02-classified forwarding delegates now repointed by WP03/WP04 (grep=0 per name) (WP18)
  • ✅ T088 Refactor the 3 production _rb.<name> call-sites to import the seam leaf directly; delete their façade delegates (WP18)
  • ✅ T089 Update the frozen test_bridge_compat_surface.py exact-baseline; surviving sentinels intact (WP18)
  • ✅ T090 Dead-code gate + tests/runtime (incl. test_bridge_parity.py) green (WP18)
  • ✅ T091 ruff/mypy clean; final runtime_bridge parity/compat tracer verdict; NFR-001 recorded (WP18)

Prompt: tasks/WP18-runtime-bridge-src-delegate-retirement.md · ~150 lines · deps: WP04