Spec Kitty

└─ kitty-specs
   └─ Test-suite remediation: masked greens, harness and pin honesty

Mission Run:

📚 Docs ↗

description: "Work package task list for test-suite-remediation-01M3SSDW"

Work Packages: Test-suite remediation: masked greens and pin honesty

Inputs: Design documents from kitty-specs/test-suite-remediation-01M3SSDW/ Prerequisites: plan.md (required), spec.md (user stories), research.md, data-model.md, contracts/dead-symbol-allowlist.md, quickstart.md, research/{masked-greens,pin-inventory,dead-symbol-rekey}.md

Tests: This mission's product under change is the test suite. Every WP edits tests, and every FIX/RETIRE carries a planted-break proof (FR-011). That test work is the deliverable itself, not optional extra work.

Organization: Fine-grained subtasks (Txxx) roll up into work packages (WPxx). Each work package is independently deliverable and testable. The plan's Implementation Concern Map (IC-01..IC-12) maps to WPs as follows:

IC-01 → WP01; IC-02 → WP02; IC-03 → WP03; IC-04 → WP04; IC-05 → WP05.
IC-06 splits along the pin-inventory G2 line (WP06) and the G1+G3 line (WP07).
IC-07 → WP08; IC-08 → WP09; IC-09 → WP10.
IC-10 splits into loader + YAML + migration (WP11) and the gate rewrite + bite battery (WP12).
IC-11 → WP13, with two re-cuts:
The refresh helper and its 17 tests are retired in WP12. test_refresh_dead_symbol_hashes.py imports _compute_dangling / _compute_offenders / _compute_stale from the gate, so it cannot outlive the gate rewrite. Retiring it in the same WP keeps every commit green.
docs/development/reference/ci-gate-mechanics.md moves to WP15, so one WP owns every docs/ edit and the regenerated docs indexes.
IC-12 splits into ratchet rows + keys floor (WP14) and ADR + docs (WP15).

Prompt Files: Each work package references a matching prompt file in tasks/. Treat this file as the high-level checklist; the deep implementation detail lives inside the prompt files.

Subtask Format: `[Txxx] [P?] Description`

[P] indicates the subtask can proceed in parallel (different files/components).
Subtasks are reference rows, not checkboxes: record completion with spec-kitty agent tasks mark-status <Txxx> --status done --mission test-suite-remediation-01M3SSDW. The reduced event-log snapshot is the sole subtask-completion authority.

Path Conventions

Single project: src/, tests/, docs/ at the repository root.
The only planned product-source change is src/specify_cli/doctrine/pack_validator.py (WP01, FR-005 fix commit only; C-005).
Any other src/ edit is allowed only as a sanctioned FR-005 fix under mission-wide rule A below: a red-first, separate fix(...) commit, with a one-line rationale in the WP notes (analysis C1).

Mission-wide rules (every WP carries them in its prompt)

C-001 / NO_FULL_HEAVY_SUITES_IN_MISSION: named test files and named gate files only. Never tests/architectural/ as a directory, never make test-full, and no stress, timing, e2e or performance suite run. A named e2e or integration file run by name counts as a named test file, not a suite (plan reading; analysis B2). make test-fast is the shared baseline.
C-011, ATDD-first (D1 reading, traces/design-decisions.md):
product-code changes take a failing-first test commit, red on base and green at the WP's final commit;
test-only WPs carry the planted-break red→green proof;
WP10 commits strict-xfail ATDD tests;
WP11 commits its loader tests failing-first, before the loader;
WP12 flips the WP10 tests green.

Every WP states this in its "Definition of Done (C-011)".

A. New product defect (FR-005, DM-01M3SSV4; analysis C1): if an unmasked or converted test exposes a new product defect, never re-mask it.
If the fix fits the WP: a red-first, separate fix(...) commit, as a sanctioned out-of-map src/ edit, with a one-line rationale in the WP notes, a new issue and an issue-matrix row.
Otherwise: a strict xfail citing a newly filed open issue, and a report to the orchestrator.
B. Evidence completeness (FR-011; analysis C2): the evidence in the review note and final report is the full per-item record (path::function::mutation, old result, new result on the break, result after revert, command), never a summary.
C. Lint and type gates (NFR-005; analysis C3): ruff check and format-check on every touched .py file; mypy on touched src/ files and on new or rewritten typed test modules (WP05, WP10, WP11, WP12). Compare against the base where findings pre-exist, and add 0 new ones.
C-007: planted breaks are scratch edits, reverted and never committed. Run git diff --stat src/ (and git diff --stat overall) before every commit.
Evidence (FR-011): records use the data-model.md §4 shape. They go into the WP's hand-off note and final report. The orchestrator materializes them under kitty-specs/.../evidence/ at closeout (see "Closeout (orchestrator)" below). Never write them into a kitty-specs/ path from a code WP.
Format-excluded files (pyproject.toml [tool.ruff.format].exclude) are edited without running ruff format on them.
CHANGELOG entries are not written by WPs; the orchestrator writes them at closeout.
Commit trailers: Co-Authored-By: Claude Sonnet 5 <[email protected]> (sonnet WPs) or Co-Authored-By: Claude Opus 5.5 <[email protected]> (opus WPs), then Claude-Session: https://claude.ai/code/session_016Yu85b3RXSx3QzQphAUzpf.

Work Package WP01: Doctrine probe unmask and fragment-intent product fix (Priority: P1) 🎯 MVP

Goal: The 9 doctrine-probe tests execute instead of skipping, and the defect that unmasking surfaced is fixed red-first: _collect_fragment_edge_intent ignores drg/fragment.yaml intent and emits a spurious same_id_collision advisory. Independent Test: uv run --frozen pytest tests/specify_cli/doctrine/test_pack_validator.py tests/integration/test_quickstart_end_to_end.py -n0 -q -rs shows 0 SKIPPED for the 9 former probe tests (was 9), and the fragment-intent regression test was red on the red commit and is green on the fix commit. Prompt: tasks/WP01-doctrine-probe-unmask-and-fragment-intent-fix.md Requirement Refs: FR-001, FR-005, FR-011, NFR-001, NFR-004, NFR-005, C-001, C-005, C-007, SC-001, SC-005

Included Subtasks

T001 Record the 9-SKIP baseline; delete the class probe in tests/specify_cli/doctrine/test_pack_validator.py and add the built_in_dir precondition assert (WP01) T002 [P] Delete the module probe in tests/integration/test_quickstart_end_to_end.py (4 tests) and add the same precondition assert (WP01) T003 Re-author the intent fixture to drg/fragment.yaml, add the validate_pack regression test, keep one *.graph.yaml case with check_drg_root=False, and file the defect issue: the RED commit (WP01) T004 Product fix in src/specify_cli/doctrine/pack_validator.py::_collect_fragment_edge_intent, in its own fix(...) commit (WP01) T005 Planted-break proofs, NFR-001 counts, skip hygiene, issue-matrix row and evidence hand-off (WP01)

Implementation Notes

Sequence: test edits (T001/T002) → red commit (T003, regression test RED on the spurious advisory) → fix commit (T004, product file only) → proofs (T005).
The precondition assert uses charter.offering.pack_paths.built_in_dir(ArtifactKind.TACTIC) / f"{_BUILT_IN_TACTIC_ID}.tactic.yaml".

Parallel Opportunities

T001 and T002 touch different files and can proceed in parallel.

Dependencies

None (starting package).

Risks & Mitigations

The fix must not alter the drg_root_graph_missing path from #3387. Run tests/doctrine/drg/test_org_fragment_validation.py by name.
The fix commit must stay product-only (C-005). Check the commit's --stat.

Estimated prompt size: ~361 lines (dense; written prompt).

Work Package WP02: Closed-issue markers: re-point, retire, re-cite (Priority: P1)

Goal: No skip, xfail or guard in the FR-002 inventory rests on a closed issue.

#3113 is re-pointed to a newly filed open issue, still strict, and the landmine guard changes in the same edit.
The #932 and EXPERIMENTAL#828 dead guards are retired, and a version-monkeypatch test replaces the unreachable branch.
The sync gate is kept and re-cited.

Independent Test: The quickstart FR-002 commands pass. Under --runxfail, both limit-8 cases FAIL with "scanner went blind to transport-call". rg -n "_has_events_5|clean_install_acceptance_deferred|shared_package_deferral" tests returns nothing. Prompt: tasks/WP02-closed-issue-markers.md Requirement Refs: FR-002, FR-005, FR-011, NFR-001, NFR-004, NFR-005, C-001, C-002, C-007, SC-002, SC-005

Included Subtasks

T006 File the "egress guard limit 8" open issue. Re-point both strict xfails, the landmine guard and the limit-8 docstring cross-reference to it in one edit (WP02) T007 [P] Delete _has_events_5 and its 6 guard sites in the 3 migration test files (WP02) T008 Replace the unreachable branch at tests/migration/test_mission_state_repair.py:161 with a version-monkeypatch refusal test (WP02) T009 [P] Retire both clean_install_acceptance_deferred() skipifs, delete tests/_support/shared_package_deferral.py, and prove the covering guards (WP02) T010 Re-cite the sync-gate docstring. Record the row-8 false positive, the skip hygiene and the evidence hand-off (WP02)

Implementation Notes

The new issue number must exist before any xfail reason is edited.
The version test monkeypatches spec_kitty_events.__version__ below REQUIRED_EVENTS_PACKAGE (5.0.0) and asserts MissionStateDryRunError matching requires spec-kitty-events >=.

Parallel Opportunities

T007 and T009 touch disjoint files.

Dependencies

None.

Risks & Mitigations

The spec-kitty#828 vs EXPERIMENTAL#828 ambiguity is stated in the evidence.
The wheel/venv tests in test_packaging_parity.py and test_clean_install_next.py are run by node, never as a suite.

Estimated prompt size: ~352 lines (dense; written prompt).

Work Package WP03: Quarantine removal and lane relocation (Priority: P1)

Goal: The five quarantined accept-diagnose tests, and the 18 other tests in their file, run in a real CI lane. The file is relocated to tests/specify_cli/acceptance/, which the nightly specify-cli-out-of-matrix job collects under -n auto --dist loadfile. The JSON parses use result.stdout. Independent Test: --collect-only under the nightly selection lists the 5 node ids. The registry check shows that tests/specify_cli/acceptance is claimed by no modules[] row. The file passes under -n 2 --dist loadfile. Prompt: tasks/WP03-quarantine-removal-and-lane-relocation.md Requirement Refs: FR-002, FR-003, FR-011, NFR-001, NFR-004, NFR-005, C-001, C-007, SC-002, SC-005

Included Subtasks

T011 git mv the file to tests/specify_cli/acceptance/test_acceptance_support.py, rename the ruff.toml per-file-ignore key, and fix the docstring path in test_issue_4891_accept_missing_lanes.py (WP03) T012 Remove _ACCEPT_COMMAND_XDIST_QUARANTINE and its 5 uses; switch every accept-CLI json.loads(result.output) to result.stdout (WP03) T013 Lane proof: --collect-only under the nightly selection, plus the registry check (RK-2) and the named architectural gates (WP03) T014 Per-test planted break, single-file xdist run and evidence hand-off (WP03)

Implementation Notes

Keep pytestmark = [pytest.mark.integration]. Keep history with git mv.
Do not touch tests/conftest.py's quarantine chokepoint (WP04 owns tests/conftest.py).

Parallel Opportunities

None inside the WP (a single file chain).

Dependencies

None.

Risks & Mitigations

RK-2: a demotion to --ignore without a registry row would orphan the file again. The registry check is recorded in the evidence.
Out of scope, and not touched (RK-5): the quarantine-visibility residue and the three open-issue quarantines.

Estimated prompt size: ~293 lines (dense; written prompt).

Work Package WP04: Errors fail instead of skipping (Priority: P2)

Goal: Build, packaging-metadata, time-budget, home-isolation and declared-dependency errors fail instead of skipping (masked-greens rows 10–15, 17, 18). Declared platform/tool guards stay exempt. Independent Test: The quickstart FR-004 named-file run is green. The --collect-only -m timing and -m stress runs each collect their test. rg -n "_build_wheel_fallback|_SKIP_PRE_WP04|importorskip\(\"jsonschema\"\)" tests returns nothing. Prompt: tasks/WP04-errors-fail-instead-of-skipping.md Requirement Refs: FR-004, FR-011, NFR-001, NFR-004, NFR-005, C-001, C-007, SC-005

Included Subtasks

T015 Rows 10–11: build_artifacts / installed_wheel_venv fail instead of skip; SharedBuildError docstring; stale #3595 citation campsite in tests/conftest.py (WP04) T016 [P] Row 12: delete the dead _build_wheel_fallback and its stale comment block in tests/doctrine/test_wheel_packaging.py (WP04) T017 [P] Row 13: convert the 11 metadata skip sites in tests/cross_cutting/versioning/test_version_detection.py to failures (WP04) T018 [P] Row 14: hoist _SC12_BUDGET_SECONDS, split the budget into a timing test that also carries stress, and prove it with --collect-only (RK-3) (WP04) T019 [P] Row 15: the home-isolation skip becomes an assert; delete _SKIP_PRE_WP04 and the stale pragmas (WP04) T020 [P] Rows 17–18: plain imports for spec_kitty_events / jsonschema; exemption-list check and evidence hand-off (WP04)

Implementation Notes

tests/conftest.py is cross-cutting: run tests/architectural/test_home_owner_behaviour.py by name.
The stress file's functional assertion texts are pinned verbatim by test_timing_coverage_invariant.py:327-337. They stay in the non-timing test.

Parallel Opportunities

T016–T020 are single-file edits and can proceed in parallel. T015 touches the two build files.

Dependencies

None.

Risks & Mitigations

RK-3: the timing marker has no live CI job. The split test carries stress through the module pytestmark, so the nightly stress job runs it. The proof is --collect-only, with no exception to C-001.

Estimated prompt size: ~348 lines (dense; written prompt).

Work Package WP05: Contract round-trip relocation map (Priority: P2)

Goal: The 10 contract round-trip cases that skip forever run through a test-side historical→canonical module relocation map. A module absent under both names fails. Independent Test: uv run --frozen pytest tests/contract/test_example_round_trip.py -n0 -q -rs shows 32 passed (29 round-trip cases plus 3 map self-tests), 0 skipped (was 19 passed, 10 skipped). Prompt: tasks/WP05-contract-round-trip-relocation-map.md Requirement Refs: FR-004, FR-011, NFR-001, NFR-004, NFR-005, C-001, C-006, C-007, SC-005

Included Subtasks

T021 Create tests/contract/_module_relocations.py with a longest-prefix-first HISTORICAL_TO_CANONICAL map and a resolver (WP05) T022 Route tests/contract/test_example_round_trip.py imports through the map; replace except ImportError: pytest.skip with a failure naming both module names (WP05) T023 Map self-test (no dead rows, every target importable), planted breaks and evidence hand-off (WP05)

Implementation Notes

Archived contracts under kitty-specs/ are immutable (C-006). The map lives test-side.

Parallel Opportunities

None (a single chain).

Dependencies

None.

Risks & Mitigations

A nested module must resolve deterministically: longest prefix first.

Estimated prompt size: ~292 lines (dense; written prompt).

Work Package WP06: #5346 pins: consolidation trio (Priority: P1)

Goal: Convert or retire #5346 rows 3, 4 and 6 (pin-inventory G2):

retire the hand-built allocator-path tests;
rename and re-scope the mission-number bake test;
stop pinning the private refusal-helper signatures.

Independent Test: The named consolidation files plus their covering guards are green. The neutral plants stay green with 0 test edits, and the violation plants go red (NFR-003). Prompt: tasks/WP06-5346-consolidation-trio.md Requirement Refs: FR-006, FR-008, FR-011, NFR-003, NFR-004, NFR-005, C-001, C-002, C-007, SC-004, SC-005

Included Subtasks

T024 [P] Row 3: retire TestWorktreeTeardownSeamRouting in tests/consolidation/test_mid8_embedded_preflight.py, with covering-guard proof (WP06) T025 [P] Row 4: rename and re-scope the unreachable-primary test in tests/consolidation/test_issue_4474_topology_aware_bake.py, and name the executor refusal guard (WP06) T026 [P] Row 6: replace full-signature assert_called_once_with pins with assert_called_once() + call_args.args[0] is exc, keeping one base_sha threading assert (WP06) T027 Neutral and violation plants per row, and the evidence hand-off (WP06)

Implementation Notes

The covering guards are read-only: tests/consolidation/test_executor_lane_naming.py, tests/consolidation/test_mission_number_truthful_4900.py and tests/lanes/test_branch_naming_seam.py.

Parallel Opportunities

T024, T025 and T026 are on three different files.

Dependencies

None.

Risks & Mitigations

RK-1: row 3 reduces the executed-test count. C-002 governs it, and the covering guard is proven red on the same break.

Estimated prompt size: ~300 lines (dense; written prompt).

Work Package WP07: #5346 pins: compat surface and copied literals (Priority: P1)

Goal: Convert or retire #5346 rows 2, 5 and 7 (pin-inventory G1 + G3):

the 18-times re-pinned len(SYMBOL_TO_MODULE) == 196;
the verbatim-copy recapture constants;
the literal-scan teardown routing test, which becomes behavioural.

Independent Test: The named files plus tests/specify_cli/cli/commands/test_mission_close_teardown_message.py are green. The neutral plants stay green with 0 test edits, and the violation plants go red. Prompt: tasks/WP07-5346-compat-surface-and-copied-literals.md Requirement Refs: FR-006, FR-008, FR-011, NFR-003, NFR-004, NFR-005, C-001, C-002, C-007, SC-004, SC-005

Included Subtasks

T028 [P] Row 2: retire test_guard_covers_full_167_symbol_surface and its changelog docstring; add the relational floor all(_SEAM_GROUPS.values()) (WP07) T029 [P] Row 5a: :481 asserts the rendered substitutions; retire :490 behind the :336/:341 guard (WP07) T030 Row 5b: :494 records the argv passed to the stubbed capture_shard_timings.main and asserts argv[:2] == ["--module", "charter"] (WP07) T031 [P] Row 7: a behavioural seam recorder test replaces the per-file literal scan in tests/specify_cli/coordination/test_teardown_single_seam_routing.py (WP07) T032 Neutral and violation plants per row, and the evidence hand-off (WP07)

Implementation Notes

test_teardown_single_seam_routing.py is format-excluded; do not reformat it.

Parallel Opportunities

The three files are disjoint, so T028, T029–T030 and T031 are parallel.

Dependencies

None.

Risks & Mitigations

"cardinality-is-contract" is residue of the retired golden-count gate (R3). Remove the marker; do not re-add it.

Estimated prompt size: ~310 lines (dense; written prompt).

Work Package WP08: FR-007 class: architectural and live-source call-site counts (Priority: P1)

Goal: Convert the exact counts of live structure in F3, F6, F8, F9, F11 and F12 (pin-inventory G4 + G5) to invariant forms or retire them behind covering guards. EXPECTED_CALL_EXPRESSION_COUNT is kept as an audited contract cardinality, with its disposition recorded. Independent Test: The IC-07 named-file run is green, and every row's neutral and violation plants behave as recorded. Prompt: tasks/WP08-fr007-architectural-and-call-site-counts.md Requirement Refs: FR-007, FR-008, FR-011, NFR-003, NFR-004, NFR-005, C-001, C-002, C-003, C-007, SC-004, SC-005

Included Subtasks

T033 [P] F9: retire test_recorded_denominator_matches_docstring_claim and strip the counts from the docstring (WP08) T034 [P] F11: a per-site partition (has_remediation or state in _PASS_STATES or (f, s) in _EXEMPT_STATES) in tests/architectural/test_remediation_effectiveness.py. The floors become >=, counted over sites (WP08) T035 [P] F12: retire EXPECTED_ENCLOSING_COUNT (set equality covers it); keep EXPECTED_CALL_EXPRESSION_COUNT as keep-contract (WP08) T036 [P] F3: floor >= 1 on materialize_calls; delete the per-site changelog docstring (WP08) T037 [P] F6: drop expected_sites from the parametrize table; add a >= 1 floor (WP08) T038 [P] F8: retire test_member_count behind :31 / :48 (WP08) T039 Neutral and violation plants per row, and the evidence hand-off (WP08)

Implementation Notes

Four of the six files are format-excluded (see the prompt).

Parallel Opportunities

Every conversion is on its own file.

Dependencies

None.

Risks & Mitigations

F11 needs a per-site partition. A pair-level one is RED on day one (5 pass-state pairs) and misses the split-site exploit (B2). The documented exploit (:348-360) and the split-site plant must both stay red.
RK-1 applies to F8, F9 and F12a.

Estimated prompt size: ~327 lines (dense; written prompt).

Work Package WP09: FR-007 class: corpus, data-artefact and agent-registry counts (Priority: P1)

Goal: Replace the counts in F4, F5, F7 and F10 (pin-inventory G6 + G7) over the shipped corpus, committed data artefacts and agent registries with shape or relational invariants. Independent Test: The IC-08 named-file run is green. The neutral plants stay green with 0 edits, and the violation plants go red. Prompt: tasks/WP09-fr007-corpus-and-registry-counts.md Requirement Refs: FR-007, FR-008, FR-011, NFR-003, NFR-004, NFR-005, C-001, C-002, C-006, C-007, SC-004, SC-005

Included Subtasks

T040 [P] F4: shape invariants in tests/specify_cli/bulk_edit/test_occurrence_map_field_paths.py; correct the "SC-011 cardinality contract" claim in place (WP09) T041 [P] F5: relational anchor invariant in tests/docs/test_glossary_linker.py; rename the _104_ test (WP09) T042 [P] F7: retire the absolute counts in tests/unit/convergence/test_census_status.py; floor >= 1 (WP09) T043 [P] F10: cross-registry set equality in tests/agent/test_agent_config_migration.py and tests/specify_cli/regression/test_twelve_agent_parity.py (WP09) T044 Neutral and violation plants per row, and the evidence hand-off (WP09)

Implementation Notes

F10 imports the canonical specify_cli.agent_utils.directories.AGENT_DIRS. m_0_9_1_complete_lane_migration exposes it only as a class attribute, so it is not the import source. The prompt explains how to avoid the tautological NON_MIGRATED_AGENTS == AGENT_COMMAND_CONFIG form.

Parallel Opportunities

All four conversions are on disjoint files.

Dependencies

None.

Risks & Mitigations

F4 has the largest docstring. Keep the per-triple contract test untouched.

Estimated prompt size: ~307 lines (dense; written prompt).

Work Package WP10: Dead-symbol ATDD contract pins and parity snapshot (Priority: P1)

Goal: Commit the acceptance contract of the re-key before the re-key lands: M1, M7, M8 and M11 against the new seam, each red on its own. Freeze the exact exempted (module, name, category) set as a parity snapshot, and record the real-tree SC-003 red observation. Independent Test: tests/architectural/test_dead_symbol_allowlist_contract.py collects cleanly. Every test is strict-XFAIL with raises=(ImportError, AttributeError), and each is red on the missing seam under --runxfail. test_no_dead_symbols.py is GREEN before the plant and RED under the code-token body-edit plant: rename the local envelope to persisted_envelope in src/specify_cli/status/lifecycle_events.py::append_lifecycle_event (:608-645). A docstring edit does not change the content-tier hash, so it must not be used. Prompt: tasks/WP10-dead-symbol-atdd-contract-pins.md Requirement Refs: FR-009, FR-011, NFR-003, NFR-004, NFR-005, C-001, C-007, SC-003, SC-005

Included Subtasks

T045 Create tests/architectural/test_dead_symbol_allowlist_contract.py: in-body seam import helper and synthetic-corpus builder (WP10) T046 M1: an allowlisted dead symbol's code-token body edit leaves offenders [] and stale [], with a same-fixture positive control (WP10) T047 M7 (a rename is reported) and M8 (a move is reported with a "probably moved to" hint) (WP10) T048 M11: an authority-parse flip over a scratch YAML; M11(a) pops from a category with ≥ 2 entries (WP10) T049 Parity snapshot script and before.json (scratchpad), the real-tree SC-003 red observation, and the evidence hand-off (WP10)

Implementation Notes

The seam import goes inside each test body, never at module scope (contract §3, ATDD rule).
The red is committed as xfail(strict=True, raises=(ImportError, AttributeError)) (orchestrator ruling). WP12 removes the markers.
before.json stays outside the repository. Its sha256, counts (293 / 91) and the script text go into the evidence hand-off, so WP12 can regenerate and verify it.

Parallel Opportunities

T046–T048 can be drafted in parallel after T045.

Dependencies

None.

Risks & Mitigations

The tests must target the contract (contracts/dead-symbol-allowlist.md), never today's SymbolKey internals, or WP12 must rewrite them.

Estimated prompt size: ~361 lines (dense; written prompt).

Work Package WP11: Dead-symbol allowlist loader and YAML migration (Priority: P1)

Goal: Add the schema-enforcing loader tests/architectural/_dead_symbol_allowlist.py (L1–L10) and the migrated data file tests/architectural/dead_symbol_allowlist.yaml, including the folded #470 widened list. Prove data-level parity against today's gate constants. The gate is not switched here; WP12 does that. Independent Test: tests/architectural/test_dead_symbol_allowlist_loader.py is green (M12 schema battery plus real-file load). The YAML carries 293 entries and 91 widened entries. Its key set equals the old allowlist's (module_path or source_module, bare_name) set. Prompt: tasks/WP11-dead-symbol-allowlist-loader-and-yaml.md Requirement Refs: FR-009, FR-011, NFR-005, C-001, C-003, C-007

Included Subtasks

T050 Loader module: DeadSymbolKey, AllowlistCategory, AllowlistEntry, WidenedEntry, DeadSymbolAllowlist, the duplicate-key-rejecting SafeLoader, rules L1–L10, ALLOWLIST (the single parse) with its views SYMBOL_ALLOWLIST and WIDENED_SCOPE_GRANDFATHERED_470 (WP11) T051 Scratchpad-only converter emitting tests/architectural/dead_symbol_allowlist.yaml: module_path wins, category_* ids, rationales, widened split, check_push_safety collapsed, tombstones dropped (WP11) T052 tests/architectural/test_dead_symbol_allowlist_loader.py: the M12 schema battery plus data-level parity against the old gate constants (WP11) T053 Quality gates (ruff, mypy, positional-anchor ban) and the evidence hand-off (WP11)

Implementation Notes

Commit order (C-011): the T052 loader test file is committed first, failing, before the T050 loader.
M12 lives in a new loader test file rather than the gate file. It exercises only the loader, and this keeps WP11 and WP12 file-disjoint.

Parallel Opportunities

T050 and T051 can be drafted in parallel. T052 needs both.

Dependencies

Depends on WP10 (ATDD-first: the contract pins land before any re-key artefact).

Risks & Mitigations

The merge_three_layers trap: module_path="charter.drg" must win over source_module="charter.offering.drg.merge".
PyYAML silently merges duplicate keys, so L3 needs the custom loader.

Estimated prompt size: ~356 lines (dense; written prompt).

Work Package WP12: Dead-symbol gate re-key and bite battery (Priority: P1)

Goal: Switch tests/architectural/test_no_dead_symbols.py to the (module, name) identity loaded from YAML:

apply the G1–G8 contract, with stale verdicts INVALID → GONE → REVIVED → SUPERSEDED → MOOT;
delete the persisted-hash data and its guards;
rebase the bite battery onto M2–M10 and M13;
migrate test_p1_planted_regression.py;
retire the refresh helper and its 17 tests, which are coupled to the gate internals being removed;
prove lossless before.json == after.json parity.

WP10's contract tests PASS once WP12 removes their strict-xfail markers, a sanctioned out-of-map edit that makes no other change to test_dead_symbol_allowlist_contract.py. The widened #470 section is evaluated from the same parsed instance through _evaluate_widened, with an M11(c) bite (F-05). Independent Test: The IC-10 named-file run is green, and the parity diff is empty. The SC-003 code-token plant (envelope → persisted_envelope in append_lifecycle_event) stays GREEN with 0 edits. A planted new dead symbol reds the gate (the C-002 covering guard for the helper retirement), and test_p1_planted_regression stays live. Prompt: tasks/WP12-dead-symbol-gate-rekey.md Requirement Refs: FR-008, FR-009, FR-011, NFR-003, NFR-005, C-001, C-002, C-003, C-007, SC-003, SC-005

Included Subtasks

T054 Before any edit, re-generate before.json with WP10's recorded script and verify its sha256 (WP12) T055 Wire the gate to the loader: _SYMBOL_ALLOWLIST / _WIDENED_SCOPE_GRANDFATHERED_470 aliases (G8), DeadSymbolKey exemption plus keyability (G1), allowlist injection (G7); keep auto-exempt condition (1) (G5) (WP12) T056 Stale verdicts (G3) with the move hint and the corpus floor. Delete the T016 suppression, dangling logic, provenance guards, cross-category duplicate guards and every _CATEGORY_* literal (G6) (WP12) T057 Rebase the bite battery: M2–M6, M9, M10 and M13. Keep bite_e, bite_k, the widened and dynamic-accessor tests; fold the disjointness check into SUPERSEDED. Retire bite_b, bite_g's body arm and bite_j's relocation arm (WP12) T058 Migrate tests/architectural/test_p1_planted_regression.py to DeadSymbolKey (WP12) T059 Retire tests/architectural/_refresh_dead_symbol_hashes.py and tests/architectural/test_refresh_dead_symbol_hashes.py (17 tests), with the C-002 covering-guard proof (WP12) T060 after.json parity, the real-tree SC-003 plant, the REVIVED plant, the contract suite green and the evidence hand-off (WP12)

Implementation Notes

The gate file keeps its name, its architectural marker and its real asserts (C-007 canon, pinned in 3 places). The real-tree test keeps its own literal assert statements (F-11).
RealTreeInputs is read-only. M13 never monkeypatches the walker before calling the cache (F-06).
Sanctioned out-of-map edits, which are not added to owned_files to avoid ownership overlap:
removing the strict-xfail markers from WP10's test_dead_symbol_allowlist_contract.py;
regenerating WP11's dead_symbol_allowlist.yaml only on base drift, with WP11's recorded converter (F-02a).
test_no_dead_symbols.py takes about 4 minutes. It is a named gate file, not a sweep.

Parallel Opportunities

T058 can proceed alongside T056–T057 once T055 lands.

Dependencies

Depends on WP11.

Risks & Mitigations

Parity drift if the base moves: re-take before.json from the new base and record both hashes.
Complexity ≤ 15 per function (NFR-005). Extract the verdict classifier into small helpers.
The size is at the upper bound (7 subtasks). All four owned files are coupled through the gate's internal API, so the WP cannot be split without shared ownership.

Estimated prompt size: ~395 lines (dense; written prompt).

Work Package WP13: Retire the hash-toll surfaces (Priority: P2)

Goal: Delete SymbolKey.source_module and its G1–G6 guard tests, which exist only for the persisted-hash identity. Correct the stale _symbol_key.py docstrings and the tests/architectural/README.md scope claim. Keep tests/unit/test_symbol_key.py's len(index) == 400. Independent Test: The IC-11 named-file run is green. rg -n "\.source_module\b|source_module=" src tests scripts and rg -n "import .*_refresh_dead_symbol_hashes|from tests\.architectural\._refresh_dead_symbol_hashes" tests scripts return nothing. The grep targets field and import usage; WP11's forbidden-key literals are expected (F-01). test_timing_coverage_invariant.py stays green. Prompt: tasks/WP13-retire-hash-toll-surfaces.md Requirement Refs: FR-009, FR-011, NFR-005, C-001, C-002, C-007, SC-005

Included Subtasks

T061 Grep gate, plus the C-002 covering-guard proof: WP11's loader rule L5 rejects a source_module key (WP13) T062 tests/architectural/_symbol_key.py: remove SymbolKey.source_module and correct the header, body-sensitivity and G1–G6 docstrings (WP13) T063 tests/unit/test_symbol_key.py: remove the G1–G6 section (25 source_module references) and keep len(index) == 400 (WP13) T064 tests/architectural/README.md stale scope claim; evidence hand-off (WP13)

Implementation Notes

_symbol_key.py is format-excluded; do not reformat it. SymbolKey itself stays, as runtime-only keyability and auto-exempt condition (1).

Parallel Opportunities

T062 and T063 are parallel after T061.

Dependencies

Depends on WP12.

Risks & Mitigations

RK-1: the deleted G-tests. Their subject (the field) is deleted, and the L5 plant is the covering guard.

Estimated prompt size: ~271 lines (dense; written prompt).

Work Package WP14: Size ratchet rows and top-level-keys floor (Priority: P1)

Goal: Cap the new dead-symbol exemption authority in _baselines.yaml (Burn-down (a)):

the allowlist_entries row and leaf;
the RK-6 widened_grandfathered_470 row and leaf, in the same section.

In the same WP:

Independent Test: tests/architectural/test_ratchet_baselines.py is green, and test_lowering_an_enforced_leaf_below_live_fails passes for both new rows. The violation plants go red. Prompt: tasks/WP14-size-ratchet-rows-and-keys-floor.md Requirement Refs: FR-006, FR-007, FR-008, FR-009, FR-010 (withdrawn; disposition record only), FR-011, NFR-003, NFR-005, C-001, C-003, C-007, C-008, SC-004, SC-005

convert len(_REQUIRED_TOP_LEVEL_KEYS) == 15 (#5346-1 / F2) to a floor at the live section count at landing (>= 16). A delete-the-new-section plant must go red (F-03);
re-plant the planted-leaf test;
record the kept-ratchet dispositions (FR-008, SC-004).

Included Subtasks

T065 Red-first: plant the new row and observe :618 red. Convert == 15 to >= <live section count> (16 today), delete the section changelog comment, and plant the section deletion, which must go red (WP14) T066 Add _SizeRatchet("test_no_dead_symbols", "allowlist_entries", "tests.architectural._dead_symbol_allowlist", "SYMBOL_ALLOWLIST") and its leaf, taken from the live count (WP14) T067 Add the RK-6 row widened_grandfathered_470 → WIDENED_SCOPE_GRANDFATHERED_470 and its leaf in the same section (WP14) T068 Re-plant test_leaf_drift_detects_planted_unenforced_leaf (:554-578) on a section that has no enforced row (WP14) T069 Violation plants, plus the keep-ratchet / keep-contract / resolved disposition records for SC-004 (FR-008 history check); evidence hand-off (WP14)

Implementation Notes

The leaf values are read from the live loader at landing, never from this plan.

Parallel Opportunities

None (one test file plus its YAML).

Dependencies

Depends on WP12.

Risks & Mitigations

C-008: no census gate is added. The rows are ordinary size ratchets.

Estimated prompt size: ~312 lines (dense; written prompt).

Work Package WP15: ADR and gate-mechanics docs (Priority: P2)

Goal: Record the identity decision in ADR docs/adr/4.x/<date>-1-dead-symbol-allowlist-module-name-identity.md, which partially supersedes D-1 of relocation-hardened-dead-code-scanners-01KX958P. Rewrite the stale "Renaming a symbol whose body is allowlisted" section of docs/development/reference/ci-gate-mechanics.md. Register the ADR and regenerate the docs indexes. Independent Test: tests/docs/test_freshen_adr_inventory.py, test_inventory_lockfile.py, test_adr_content_invariance.py and tests/architectural/test_no_legacy_terminology.py are green, and scripts/docs/check_docs_freshness.py --ci passes. Prompt: tasks/WP15-adr-and-gate-mechanics-docs.md Requirement Refs: FR-009, FR-011, NFR-005, C-001, C-006, C-008

Included Subtasks

T070 Author the ADR: context, decision, consequences, supersedes / relates, non-goals (WP15) T071 Rewrite the "Renaming a symbol whose body is allowlisted" section of docs/development/reference/ci-gate-mechanics.md, and add the dead-symbol YAML kind to docs/development/how-to/add-architectural-gate-exemption.md (F-13) (WP15) T072 Register the ADR with python -m scripts.docs.freshen_adr_inventory (index row + page-inventory lockfile) (WP15) T073 Regenerate the docs retrieval index and inventory lockfile, run the freshness check and the named docs tests, and hand off (WP15)

Implementation Notes

WP15 owns every docs/ edit in this mission and the regenerated indexes. No other WP touches docs/.

Parallel Opportunities

T070 and T071 are parallel.

Dependencies

Depends on WP13 and WP14 (the ADR cites the retired helper and the size-ratchet rows as landed facts).

Risks & Mitigations

The ADR filename date is the landing date. owned_files uses a date-agnostic glob.

Estimated prompt size: ~298 lines (dense; written prompt).

Dependency & Execution Summary

WP01  WP02  WP03  WP04  WP05  WP06  WP07  WP08  WP09     (independent, parallel lanes)
WP10 ──► WP11 ──► WP12 ──► WP13 ──┐
                      └──► WP14 ──┴──► WP15

Sequence: WP01–WP09 have no dependencies and are file-disjoint. The dead-symbol chain is the only sequence: WP10 → WP11 → WP12 → {WP13, WP14} → WP15.
Parallelization: Up to nine lanes for WP01–WP09, plus the dead-symbol chain in its own lane. WP13 and WP14 can run in parallel once WP12 is approved.
MVP Scope: WP01 (the unmask plus the one real product defect) and WP10–WP12 (the dominant friction source, SC-003).
Model tiers:
WP10, WP11, WP12 and WP15 run on claude-opus-5-5 (the re-key and the ADR).
All others run on claude-sonnet-5.
Review runs on opus for every WP (plan).

Closeout (orchestrator)

These steps belong to the orchestrator, not to any WP. They run after all WPs are approved, in this order.

1. Materialize the evidence (M7). Write every WP's reported evidence records into kitty-specs/test-suite-remediation-01M3SSDW/evidence/IC-NN-<slug>.md (data-model §4), and WP10's and WP12's parity JSONs and digests into evidence/dead-symbol-parity/{before,after}.json, on the planning branch (a planning artifact).

2. Rebase re-port (F-02b). After rebasing onto upstream main, check whether any upstream commit edited the (now deleted) _CATEGORY_* allowlist literals in tests/architectural/test_no_dead_symbols.py. If one did:

3. Consolidated-branch gate run (F-04, F-02b). On the consolidated mission branch, run the named files tests/architectural/test_no_dead_symbols.py, test_dead_symbol_allowlist_loader.py, test_dead_symbol_allowlist_contract.py and test_ratchet_baselines.py. The chain's parity was measured on a lane without WP01's product change. These are named gate files, not a directory sweep. 4. Follow-up issues plus issue-matrix rows (m13, RK-2, RK-5, D-14). For each, run spec-kitty agent issue-verdict --verdict deferred-with-followup. First confirm that issue-verdict creates a row when the mission has no issue matrix yet, since WP01 T005 and WP02 T010 rely on it:

5. Tracker hygiene (Standing Order 8; analysis C4). #5346 and #5353 are already assigned and commented, and their issue-matrix rows exist. At closeout, verify that their verdicts are terminal: #5346 fixed, and the parent #5353 linked with its verdict. Correct them with spec-kitty agent issue-verdict if they are not. 6. CHANGELOG entries (the WPs write none).

Split concerns merge into one IC file (analysis C5): IC-06 ← WP06 + WP07; IC-10 ← WP11 + WP12; IC-11 ← WP12 (refresh-helper retirement) + WP13; IC-12 ← WP14 + WP15. Every other IC maps to one WP.
SC-005 owner (analysis C6): each per-WP reviewer re-runs its WP's Review-Guidance picks during review and reports which ones. At closeout the orchestrator tallies them (≥ 6 FIX and ≥ 2 RETIRE across the mission), and sets reviewer_rerun: true on those records.
Update the quickstart FR-009 parity diff so it points at the committed files. If they are not committed, compare the WP10 and WP12 recorded digests instead.
port its delta into tests/architectural/dead_symbol_allowlist.yaml (with WP11's converter rules: module_path wins, category_* ids);
re-read the live counts for the two _baselines.yaml test_no_dead_symbols leaves;
re-run the parity check.
the quarantine-visibility lane residue;
the three open-issue quarantines (EXP#1021 ×2, EXP#901);
the tracker-slice importorskip("spec_kitty_tracker.context") dead guards;
the inline_meta_read three-authority count;
simplifying auto-exempt condition (1);
optionally, re-keying F12's EXPECTED_CALL_EXPRESSION_COUNT as {qualname: n};
the RK-2 "collected by some lane" gate;
the stale product comment in src/specify_cli/upgrade/migrations/m_unify_charter_activation_finalize.py:474-484 (F-15; C-005 forbids editing it here);
the project CLAUDE.md line "Canonical source: m_0_9_1_complete_lane_migration → AGENT_DIRS", which should name specify_cli.agent_utils.directories (analysis D3; outside the file set).

Requirements Coverage Summary

Requirement ID	Covered By Work Package(s)
FR-001	WP01
FR-002	WP02, WP03
FR-003	WP03
FR-004	WP04, WP05
FR-005	WP01, WP02
FR-006	WP06, WP07, WP14
FR-007	WP08, WP09, WP14
FR-008	WP06, WP07, WP08, WP09, WP12, WP14
FR-009	WP10, WP11, WP12, WP13, WP14, WP15
FR-010	Withdrawn (DM-01M3SVDP). Mapped to WP14 only because the finalize-tasks coverage check requires every FR to be mapped; WP14 records the withdrawn disposition and implements no gate (C-008)
FR-011	WP01–WP15
NFR-001	WP01, WP02, WP03, WP04, WP05 (masked-green files only, RK-1)
NFR-002	Withdrawn; no WP
NFR-003	WP06, WP07, WP08, WP09, WP10, WP12, WP14
NFR-004	WP01–WP10
NFR-005	WP01–WP15
C-001	WP01–WP15
C-002	WP02, WP06, WP07, WP08, WP09, WP12, WP13
C-003	WP08, WP11, WP12, WP14
C-005	WP01
C-006	WP05, WP09, WP15
C-007	WP01–WP14
C-008	WP14, WP15
SC-001	WP01
SC-002	WP02, WP03
SC-003	WP10, WP12
SC-004	WP06, WP07, WP08, WP09, WP14
SC-005	WP01–WP10, WP12, WP13, WP14
SC-006	Withdrawn; no WP

C-004 (the scope boundary, amended for analysis F2) is honoured. No WP runs a slice-level quality pass of the integration harness or of the status, agent, tracker or auth slices. Individually inventoried FR-006/FR-007 pins stay in scope wherever their files live, e.g. WP09's F10 edit in tests/agent/test_agent_config_migration.py.

SC-004 tally note (analysis A1): F1 and F2 are the same pins as #5346-2 and #5346-1. Count each once, under its #5346 row.

Subtask Index (Reference)

Subtask ID	Summary	Work Package	Priority	Parallel?
T001	Pack-validator probe → precondition assert	WP01	P1	Yes
T002	Quickstart e2e probe → precondition assert	WP01	P1	Yes
T003	fragment.yaml fixture + regression test (RED commit) + new issue	WP01	P1	No
T004	`_collect_fragment_edge_intent` product fix commit	WP01	P1	No
T005	WP01 plants, counts, hygiene, evidence	WP01	P1	No
T006	#3113 re-point + landmine guard + docstring	WP02	P1	No
T007	Delete `_has_events_5` ×3 and 6 guard sites	WP02	P1	Yes
T008	Version-monkeypatch refusal test	WP02	P1	No
T009	Retire #828 skipifs + delete deferral helper	WP02	P1	Yes
T010	Sync-gate re-cite, false positive, evidence	WP02	P1	No
T011	`git mv` + ruff.toml key + docstring path	WP03	P1	No
T012	Remove quarantine + `result.stdout`	WP03	P1	No
T013	Lane proof + registry check	WP03	P1	No
T014	WP03 plant + xdist + evidence	WP03	P1	No
T015	Build fixtures fail instead of skip + #3595 campsite	WP04	P2	No
T016	Delete `_build_wheel_fallback`	WP04	P2	Yes
T017	Version-detection 11 skip sites → fail	WP04	P2	Yes
T018	Stress budget → `timing`+`stress` test	WP04	P2	Yes
T019	Home-isolation skip → assert	WP04	P2	Yes
T020	Plain imports + evidence	WP04	P2	Yes
T021	`_module_relocations.py` map + resolver	WP05	P2	No
T022	Round-trip through the map; fail on absence	WP05	P2	No
T023	Map self-test + plants + evidence	WP05	P2	No
T024	#5346-3 retire allocator-path class	WP06	P1	Yes
T025	#5346-4 rename + re-scope	WP06	P1	Yes
T026	#5346-6 signature pins → identity asserts	WP06	P1	Yes
T027	WP06 plants + evidence	WP06	P1	No
T028	#5346-2 retire count + relational floor	WP07	P1	Yes
T029	#5346-5 `:481` convert, `:490` retire	WP07	P1	Yes
T030	#5346-5 `:494` argv through the real seam	WP07	P1	No
T031	#5346-7 behavioural seam recorder	WP07	P1	Yes
T032	WP07 plants + evidence	WP07	P1	No
T033	F9 retire denominator test	WP08	P1	Yes
T034	F11 partition invariant	WP08	P1	Yes
T035	F12 retire ENCLOSING, keep CALL_EXPRESSION	WP08	P1	Yes
T036	F3 floor ≥ 1	WP08	P1	Yes
T037	F6 drop `expected_sites`	WP08	P1	Yes
T038	F8 retire member count	WP08	P1	Yes
T039	WP08 plants + evidence	WP08	P1	No
T040	F4 shape invariants	WP09	P1	Yes
T041	F5 relational anchors + rename	WP09	P1	Yes
T042	F7 retire absolute counts	WP09	P1	Yes
T043	F10 registry set equality	WP09	P1	Yes
T044	WP09 plants + evidence	WP09	P1	No
T045	Contract test scaffold	WP10	P1	No
T046	M1 body edit stays green	WP10	P1	Yes
T047	M7 rename + M8 move reported	WP10	P1	Yes
T048	M11 authority parse	WP10	P1	Yes
T049	Parity snapshot + SC-003 red + evidence	WP10	P1	No
T050	Loader module L1–L10	WP11	P1	Yes
T051	Scratch converter → YAML	WP11	P1	Yes
T052	Loader tests (M12) + data parity	WP11	P1	No
T053	WP11 quality gates + evidence	WP11	P1	No
T054	Re-take `before.json`	WP12	P1	No
T055	Wire gate to loader (G1/G5/G7/G8)	WP12	P1	No
T056	Stale verdicts + deletions + corpus floor	WP12	P1	No
T057	Bite battery rebase	WP12	P1	No
T058	`test_p1_planted_regression.py` key migration	WP12	P1	Yes
T059	Retire refresh helper + 17 tests (C-002 proof)	WP12	P1	No
T060	`after.json` parity + plants + evidence	WP12	P1	No
T061	Grep gate + C-002 proof (L5 plant)	WP13	P2	No
T062	`SymbolKey.source_module` removal	WP13	P2	Yes
T063	`test_symbol_key.py` refs (keep `== 400`)	WP13	P2	Yes
T064	README scope claim + evidence	WP13	P2	No
T065	`== 15` → floor, red-first	WP14	P1	No
T066	`allowlist_entries` row + leaf	WP14	P1	No
T067	`widened_grandfathered_470` row + leaf	WP14	P1	No
T068	Re-plant the planted-leaf test	WP14	P1	No
T069	Plants + kept-ratchet dispositions + evidence	WP14	P1	No
T070	Author the ADR	WP15	P2	Yes
T071	Rewrite the gate-mechanics section	WP15	P2	Yes
T072	Register the ADR (freshener)	WP15	P2	No
T073	Regenerate docs indexes + checks	WP15	P2	No

Spec Kitty

Work Packages: Test-suite remediation: masked greens and pin honesty

Subtask Format: [Txxx] [P?] Description

Path Conventions

Mission-wide rules (every WP carries them in its prompt)

Work Package WP01: Doctrine probe unmask and fragment-intent product fix (Priority: P1) 🎯 MVP

Included Subtasks

Implementation Notes

Parallel Opportunities

Dependencies

Risks & Mitigations

Work Package WP02: Closed-issue markers: re-point, retire, re-cite (Priority: P1)

Included Subtasks

Implementation Notes

Parallel Opportunities

Dependencies

Risks & Mitigations

Work Package WP03: Quarantine removal and lane relocation (Priority: P1)

Included Subtasks

Implementation Notes

Parallel Opportunities

Dependencies

Risks & Mitigations

Work Package WP04: Errors fail instead of skipping (Priority: P2)

Included Subtasks

Implementation Notes

Parallel Opportunities

Dependencies

Risks & Mitigations

Work Package WP05: Contract round-trip relocation map (Priority: P2)

Included Subtasks

Implementation Notes

Parallel Opportunities

Dependencies

Risks & Mitigations

Work Package WP06: #5346 pins: consolidation trio (Priority: P1)

Included Subtasks

Implementation Notes

Parallel Opportunities

Dependencies

Risks & Mitigations

Work Package WP07: #5346 pins: compat surface and copied literals (Priority: P1)

Included Subtasks

Implementation Notes

Parallel Opportunities

Dependencies

Risks & Mitigations

Work Package WP08: FR-007 class: architectural and live-source call-site counts (Priority: P1)

Included Subtasks

Implementation Notes

Parallel Opportunities

Dependencies

Risks & Mitigations

Work Package WP09: FR-007 class: corpus, data-artefact and agent-registry counts (Priority: P1)

Included Subtasks

Implementation Notes

Parallel Opportunities

Dependencies

Risks & Mitigations

Work Package WP10: Dead-symbol ATDD contract pins and parity snapshot (Priority: P1)

Included Subtasks

Implementation Notes

Parallel Opportunities

Dependencies

Risks & Mitigations

Work Package WP11: Dead-symbol allowlist loader and YAML migration (Priority: P1)

Included Subtasks

Implementation Notes

Parallel Opportunities

Dependencies

Risks & Mitigations

Work Package WP12: Dead-symbol gate re-key and bite battery (Priority: P1)

Included Subtasks

Implementation Notes

Parallel Opportunities

Dependencies

Risks & Mitigations

Work Package WP13: Retire the hash-toll surfaces (Priority: P2)

Included Subtasks

Implementation Notes

Subtask Format: `[Txxx] [P?] Description`