Mission Specification: Tasks Degod Wave 2: Render Seam + Relocation
Mission Branch: degod-follow-ups (mission tasks-py-degod-wave2-01KWH9EQ, coord branch kitty/mission-tasks-py-degod-wave2-01KWH9EQ) Created: 2026-07-02 Status: Draft (post-spec squad findings folded 2026-07-02) Input: Issue #2305 handover debrief (Wave 2 of the tasks.py degod program, parent epic #2173) + live re-census against merged upstream/main 381db8d5f, plus domain-matched boyscouting from #2034. A 4-lens post-spec adversarial squad (sizing / fakeability / code-truth / structure) corrected the census and hardened the DoDs; see post-spec-squad-findings.md.
Context
Wave 1 (mission tasks-py-degod-01KWF08S, PR #2303, merged as 381db8d5f) decomposed the agent tasks god-command's decision logic into pure tested cores behind injected ports and thinned every fat command body — behavior byte-identical, guarded by the CLI contract harnesses. It deliberately deferred two goals to this mission (spec Deferred section): the Render-seam unification and the whole-file shim relocation with its LOC ceiling. During Wave 1, tasks.py grew to 4569 LOC: decision logic left for the cores while the orchestrators, glue helpers, and port-seam adapter classes accumulated in the file.
Re-census facts (merged main, 2026-07-02, squad-verified) — supersede all debrief line references and the pre-squad draft of this spec:
print(json.dumps(...)) sites (lines 508, 546, 559, 2477, 2805, 3349–3350 (one multiline call), 3474, 3488, 3585, 3665, 3863, 4557) plus one return json.dumps(payload, indent=2) status leg at line 1235. No aliased from json import dumps forms exist today.
_do_finalize_tasks (3122), _do_map_requirements (3677), _do_status (4431).
9 _ms_, and 4 _ft_ (finalize — _ft_resolve_context/_ft_validate/ _ft_apply_writes/_ft_output).
_FinalizeState (2924), _MapReqState (3188), _StatusState (3893).
_default_status_ports, _default_mark_status_ports, _default_finalize_ports.
_StatusRender (1222), _MarkStatusCoordRouter (2356).
e.g. _output_result/_output_error (56 in-file call sites), _find_mission_slug (10 sites, patched 66× in tests), _ensure_target_branch_checked_out (10 sites, patched 50×), _coord_topology_active, _skip_target_branch_commit, resolve_primary_branch, _validate_ready_for_review (patched 12×), _check_unchecked_subtasks (patched 12×).
--help fixtures asserted byte-exact; the JSON legs are shape-checked (_shape() collapses values/ordering), NOT byte-checked.
including T004/T005: the #2300 skip-vs-refuse divergence pins (move_task coord skip-exit-0 arm with wrong-leg detector; mark_status/map_requirements refuse-exit-1 arm), and the branch-coverage ratchet _BRANCH_COVERAGE_FLOORS (move_task 65% / status 48% / map_requirements 46%) whose _mutating_function_line_ranges() walks tasks.py's AST by function name — it is line-range-coupled to the fat wrappers living in tasks.py.
placement is deliberate and docstring-recorded: non-CLI consumers import without loading the agent Typer package); src/specify_cli/cli/commands/agent/tasks_ports.py is a 7-line re-export shim.
(573), tasks_status_view.py (233), tasks_mapping_core.py (156).
distinct symbols (367 by grep), plus 37 monkeypatch.setattr targeting tasks and 121 from ...agent.tasks import statements in tests. Highest-traffic symbols: locate_project_root ×67, _find_mission_slug ×66, _ensure_target_branch_checked_out ×50, get_mission_type ×26, feature_status_lock ×23, get_main_repo_root ×19. Orchestrators call these by bare name today — a module-level re-export alone does NOT preserve patch interception after relocation (see FR-002/Domain Language "seam bridge").
parent-module attribute routing** (from ...agent import mission as _mission inside functions; all infra calls go through _mission.<attr> — 40+ occurrences across its sibling modules, no module-level back-import, no cycle). Its final shim registers command callables via app.command(name=...)(fn); siblings declare no __all__ and pass the dead-symbol gate.
(imported at 150, called at 1367 inside _mt_resolve_targets).
26,612 collected tests are selected by no marker gate repo-wide. The tasks domain is already near-clean: all 39 tasks-domain test files carry pytestmark; ~1 file is gate-invisible ([unit, git_repo] without fast). The pre-existing orphan ratchet tests/architectural/test_gate_coverage.py freezes ~9.8k orphans in _gate_coverage_baseline.json (refreshable via --update-baseline — a widening risk FR-009 must close for this domain).
src/specify_cli/cli/commands/agent/tasks.py= 4569 LOC, 112 top-level defs.- JSON emission sites: 13 total distinct call sites — 12 compact
- Symbol taxonomy (the honest relocation inventory):
- 5
_do_<cmd>orchestrators:_do_move_task(2082),_do_mark_status(2641), - 61 family glue helpers across FIVE families: 23
_mt_, 11_mr_, 14_st_*, - 5 per-family State dataclasses:
_MoveTaskState(1249),_MarkStatusState(2325), - 5 per-family port factories:
_default_move_task_ports,_default_map_requirements_ports, - 4 port-seam adapter classes:
_MoveTaskCoordRouter(1120),_MapReqCoordRouter(1172), - ~30 genuinely cross-family shared helpers with no home in a per-family split —
- Contract harnesses (the parity guard is TWO files, 43 cases combined):
tests/specify_cli/cli/commands/agent/test_tasks_cli_contract.py— 27 cases: 10tests/specify_cli/cli/commands/agent/test_tasks_cli_contract_coord.py— 16 cases,- Ports content lives in
src/specify_cli/agent_tasks_ports.py(381 LOC — top-level - Pure cores (all at
src/specify_cli/cli/commands/agent/):tasks_transition_core.py - Patch-seam surface: ~370
patch("...agent.tasks.<sym>")call sites across ~40 - The
mission.pydegod template solved the seam-bridge problem with **lazy - A self-review-fallback precedence guard (
_self_review_fallback_option_error) exists - #2034 residual: no
ci-quality.ymlgate selects-m unitor-m contract; 257 of
User Scenarios & Testing (mandatory)
User Story 1 - tasks.py becomes a true registration shim (Priority: P1)
A maintainer opening src/specify_cli/cli/commands/agent/tasks.py finds only thin @app.command registration wrappers (plus the four already-small bodies and the deliberate seam-bridge/re-export surface). The five command families' orchestrators, glue helpers, State dataclasses, and port factories live in focused sibling modules; the cross-family shared helpers live in a named shared module; the port-seam adapter classes live in a dedicated adapters module. Navigating, reviewing, and changing one command no longer requires scrolling a 4569-line file, and the whole-file size is enforced by an architectural gate so the god-file cannot silently regrow.
Why this priority: This is the mission's charter — the whole-file shim-relocation goal deferred from Wave 1. The file's size is the remaining structural debt; the per-move risk is low but the volume (~3150 LOC) demands its own reviewed mission.
Independent Test: After relocation, both contract harnesses (43 cases) pass unmodified, the byte-freeze suite (see US2) is green, the full tasks-domain test surface passes with patch interception preserved, and the whole-file LOC gate passes at the recorded ceiling; a synthetic violation (a dummy function pushing the file over the ceiling) turns the gate red.
Acceptance Scenarios:
1. Given merged main 381db8d5f behavior pinned by the two contract harnesses and the byte-freeze suite, When all five command families, the shared helpers, and the adapter classes are relocated to sibling modules, Then all 43 harness cases and every byte-freeze case pass unmodified, and the ~370 patch-seam call sites still intercept the production call path (not merely resolve). 2. Given the relocated layout, When the architectural LOC gate runs, Then tasks.py is at or under the recorded ceiling, the ceiling is min(achieved, 1400), and the rationale records the delta from the 4569 baseline. 3. Given a future change re-adding orchestration bulk to tasks.py, When CI runs, Then the LOC gate fails (proven non-vacuous by a self-mutation test). 4. Given each relocation WP, When an orchestrator moves out of tasks.py, Then the branch-coverage ratchet in the coord harness is re-pointed to the relocated function in the same WP — never deleted, never floor-lowered.
User Story 2 - One rendering authority for command output (Priority: P2)
A maintainer changing how an agent tasks subcommand emits JSON touches exactly one seam: the Render port. No inline json.dumps remains in the tasks command surface; the single indent=2 status leg is served by the same parameterized seam (the _StatusRender subclass override collapses). An AST-based gate keeps the count at zero across the whole command-surface directory, including aliased and rebound forms.
Why this priority: This is the render-seam unification deferred from Wave 1 — the remaining split-brain between the Render port and 13 ad-hoc emission sites. Smaller than Stream B but it closes a rival-seam class (#2173 discipline: one production adapter per port).
Independent Test: The byte-freeze suite (one byte-exact case per emission site, frozen BEFORE any routing change) stays green through the seam unification; an AST census shows 0 inline json.dumps in the command surface; the AST gate goes red on each synthetically inserted evasion form.
Acceptance Scenarios:
1. Given byte-freeze characterization tests capturing the exact current bytes of all 13 emission sites (12 compact + the indent=2 leg) committed and green BEFORE any routing change, When the 12 compact sites are routed through Render.json_envelope, Then every byte-freeze case still passes (compact separators preserved). 2. Given the indent=2 status leg (line 1235), When the Render seam gains a parameterized indent (or an indented-envelope capability), Then the status byte-freeze case still passes and the _StatusRender(RealRender) override is deleted. 3. Given a new code path emitting JSON via json.dumps(...), from json import dumps, import json as <alias>; <alias>.dumps(...), or a rebound local (d = json.dumps; d(...)), When the AST gate runs, Then it fails with an actionable message naming the offending site — each form has its own self-mutation proof.
User Story 3 - Tasks-domain tests are CI-visible (boyscout, Priority: P3)
A contributor adding a test to the tasks domain cannot author it into CI-invisibility: every file matching the committed tasks-domain glob carries CI-selected markers, the orphan-ratchet baseline cannot re-absorb them, and the upstream marker-gap issue (#2034) is refreshed with the 2026-07-02 re-census so the repo-wide fix can be planned on current facts.
Why this priority: Domain-matched campsite cleaning (charter standing order 2). The mission adds new test files and gates; authoring them into an invisible marker class would be self-defeating. The squad found the domain near-clean already (~1 invisible file), so this is honestly small. The repo-wide structural CI fix stays out of scope.
Independent Test: A committed marker-census artifact lists every file matching the tasks-domain glob with the CI gate that selects it (zero unselected); the orphan-ratchet baseline contains no entry under the glob; #2034 carries the refreshed census comment.
Acceptance Scenarios:
1. Given the committed tasks-domain glob (see FR-009), When the marker census runs, Then every matching test file (existing + added by this mission) is selected by at least one CI gate, evidenced by a committed census artifact. 2. Given the orphan ratchet _gate_coverage_baseline.json, When the mission's test files land, Then no path under the tasks-domain glob is added to the baseline (absorbing them via --update-baseline is a violation, not a fix). 3. Given the 2026-07-02 re-census, When the mission reaches review, Then #2034 has a comment with the refreshed facts and recommended structural options.
Edge Cases
calling infra by bare name resolves in the SIBLING's namespace — @patch("...agent.tasks.<sym>") no longer intercepts even though a re-export keeps the name importable; defensively-patched tests (no assert_called) then pass green while real side effects fire. The seam bridge (FR-002) exists precisely for this; "the tests pass" is not sufficient evidence.
move_task/status/map_requirements line ranges from tasks.py's AST by name; thinning those wrappers collapses the measured range → red on a correct move. The remedy is re-pointing the ratchet target in the same WP (FR-012) — deleting or floor-lowering it is forbidden.
needed by command modules — the dedicated tasks_command_adapters.py breaks the cycle; any deviation records why no cycle arises. The lazy parent-module routing keeps family siblings acyclic (no module-level back-import).
must be sequenced before or with the status-family relocation so the class is deleted, not moved-then-deleted (a wasted move + a second seam re-point).
botched move — the move is reverted and re-done. Re-pointing the coverage ratchet's AST target (FR-012) is NOT fixture adjustment; changing an expected-output fixture to absorb a behavior diff IS, and is forbidden.
must not trip the gate (call-node inspection, not text); evasion forms (alias, rebind, move-next-door into an unscanned module) must (directory-glob scope + per-form self-mutation cases).
If the honest relocated size exceeds 1400, that is an escalation to the operator with the delta-from-4569 analysis — never a self-certified re-baseline.
the venv must match uv.lock before running or re-freezing anything (Wave 1 trap).
issue-matrix, and review artifacts the gates read live on the coordination branch, not the primary checkout (Wave 1 close friction).
- Patch interception loss (the dominant relocation hazard): a relocated orchestrator
- Branch-coverage ratchet false-red: the coord harness ratchet resolves
- Import cycle from adapter relocation: adapters subclass port classes AND are
_StatusRenderrelocate-then-delete rework: the render-seam work (FR-005/FR-006)- Golden delta on a "pure" move: any contract-harness or byte-freeze delta means a
- AST gate false positives/negatives: docstring mentions of
json.dumps(line 1225) - LOC ceiling honesty: the recorded enforcement ceiling is
min(achieved, 1400). - Typer/rich version skew: contract
--helpfixtures are typer-version-coupled — - Coord-authority writes: this is a
topology: coordmission — acceptance-matrix,
Domain Language
delegate to relocated orchestrators, plus the deliberate seam-bridge/re-export surface.
port factory. Five families: move_task (_mt_), map_requirements (_mr_), status (_st_), mark_status (_ms_), finalize_tasks (_ft_*).
serving multiple command families; relocates to the shared module, never duplicated per family.
INTERCEPTING after relocation. Default (the mission.py template mechanism): relocated code performs a lazy in-function from specify_cli.cli.commands.agent import tasks as _tasks and calls infra via _tasks.<attr>, so the patched module attribute IS the called object. "Re-point the patches" is reserved for symbols that are themselves defined-and-relocated (their agent.tasks.* patch target disappears). A bare module-level re-export alone is NOT a seam bridge.
Render.human); one production adapter serves all sites (indent parameterized).
test_tasks_cli_contract_coord.py (16 cases, incl. the #2300 divergence pins T004/T005 and the branch-coverage ratchet). 43 cases combined.
asserting the exact stdout bytes of each of the 13 JSON emission sites — the byte-level parity contract the shape-checked harness legs do not provide.
Terminology Canon binds all moved and new code).
- Registration shim: a command module containing only
@app.commandwrappers that - Command family: one subcommand's orchestrator + glue helpers + State dataclass +
- Shared helper: a cross-family helper (e.g.
_output_result,_find_mission_slug) - Seam bridge: the mechanism that keeps
@patch("...agent.tasks.<sym>")tests - Render seam / Render port: the injected output boundary (
Render.json_envelope/ - Contract harnesses:
test_tasks_cli_contract.py(27 cases) + - Byte-freeze suite: NEW characterization tests (this mission, FR-005 pre-step)
- Gate-visible: selected by at least one CI marker/path gate in
ci-quality.yml. - Canonical term is Mission;
--missionis the only primary CLI flag form (charter
Requirements (mandatory)
Functional Requirements
| ID | Title | User Story | Priority | Status |
|---|---|---|---|---|
| FR-001 | Relocate all five command families out of tasks.py into per-family sibling modules under src/specify_cli/cli/commands/agent/ — each family's move-set is its orchestrator + its glue helpers + its State dataclass + its _default__ports factory (move_task: _do_move_task + 23 _mt_ + _MoveTaskState + factory; map_requirements: _do_map_requirements + 11 _mr_ + _MapReqState + factory; status: _do_status + 14 _st_ + _StatusState + factory; mark_status: _do_mark_status + 9 _ms_ + _MarkStatusState + factory; finalize_tasks: _do_finalize_tasks + 4 _ft_ + _FinalizeState + factory). | US1 | High | Open |
| FR-002 | Preserve patch interception across every relocation via the seam bridge: relocated code routes infra calls through lazy parent-module attribute access (_tasks.<attr>, the mission.py template mechanism) by default; symbols that are themselves defined-and-relocated get their test patches re-pointed in the same WP. Each relocation WP carries the affected symbol×patch-count inventory (from the ~40-symbol census) as its seam checklist, plus a positive interception check (the patched attribute is the object production calls) for defensively-patched seams. | US1 | High | Open |
| FR-003 | Relocate the ~30 cross-family shared helpers (led by _output_result/_output_error, _find_mission_slug, _ensure_target_branch_checked_out) into a named shared sibling module with the same seam-bridge obligations; no helper is duplicated per family; no shared helper remains in tasks.py unless it is part of the registration surface itself. | US1 | High | Open |
| FR-004 | Relocate the port-seam adapter classes (_MoveTaskCoordRouter, _MapReqCoordRouter, _MarkStatusCoordRouter — and _StatusRender only if FR-006 has not already deleted it) into a dedicated tasks_command_adapters.py, with a recorded no-cycle argument for any deviation. | US1 | High | Open |
| FR-005 | Byte-freeze first, then route: commit byte-exact characterization tests for all 13 JSON emission sites (12 compact + the indent=2 leg) against the pre-change tree, THEN route the 12 compact sites through the Render port with the byte-freeze suite green throughout. | US2 | High | Open |
| FR-006 | Parameterize the Render seam's indent (or add an indented-envelope capability) so the indent=2 status leg uses the generic seam and the _StatusRender(RealRender) override is deleted; the status byte-freeze case stays green. | US2 | High | Open |
| FR-007 | Add an AST-based architectural gate asserting 0 inline json.dumps calls across the command-surface directory glob (src/specify_cli/cli/commands/agent/ including all new sibling/adapter modules), covering all evasion forms — attribute call, from json import dumps, module alias, and local rebinding — with a self-mutation proof per form. | US2 | High | Open |
| FR-008 | Reduce tasks.py to thin @app.command registration wrappers, the four already-small bodies (list_tasks, add_history, validate_workflow, list_dependents), and the deliberate seam-bridge/re-export surface; decide and execute the disposition of the 7-line cli/commands/agent/tasks_ports.py re-export shim with rationale recorded (full unshim of other files stays with #2289–#2293). New sibling modules follow the template precedent (no __all__ required — the charter's __all__ convention binds src/charter/ and src/kernel/ only). | US1 | High | Open |
| FR-009 | Make the tasks-domain test surface gate-visible with non-gameable evidence: commit the tasks-domain glob (tests/tasks/*, tests/specify_cli/cli/commands/agent/test_tasks, plus every test file this mission adds), produce a committed marker-census artifact mapping each matching file to its selecting CI gate (zero unselected), fix the ~1 invisible file, and assert that no path under the glob is present in (or added to) _gate_coverage_baseline.json. | US3 | Medium | Open |
| FR-010 | Refresh #2034 upstream with the 2026-07-02 re-census (257/26,612 marker-invisible repo-wide; original failure list largely fixed/re-marked; current gate inventory) and the recommended structural options — the repo-wide fix itself is out of scope. (Initial refresh comment posted at spec time; final census posted at review.) | US3 | Medium | Open |
| FR-011 | Add a tests/architectural/ whole-file LOC gate for tasks.py: enforcement ceiling = min(achieved, 1400); the gate ships with a non-vacuity self-mutation test; the recorded rationale states the achieved LOC and the delta from the 4569 baseline. An achieved size above 1400 is an operator escalation, never a self-certified re-baseline. | US1 | High | Open |
| FR-012 | Re-point the coord-harness branch-coverage ratchet (_BRANCH_COVERAGE_FLOORS / _mutating_function_line_ranges) to each relocated orchestrator in the same WP as its move; deleting the ratchet, lowering its floors, or leaving it measuring the thinned wrapper is forbidden. Ratchet re-pointing is expressly NOT "fixture adjustment". | US1 | High | Open |
Non-Functional Requirements
| ID | Title | Requirement | Category | Priority | Status |
|---|---|---|---|---|---|
| NFR-001 | Behavior parity, honestly defined | At every commit of every WP: (a) all 43 contract-harness cases pass unmodified; (b) the 10 --help fixtures remain byte-exact; (c) the byte-freeze suite (13 cases, FR-005) remains byte-exact; (d) expected-output fixtures are never adjusted to absorb a diff (revert-and-redo instead); (e) ratchet re-pointing per FR-012 is the sole sanctioned harness edit. | Correctness | High | Open |
| NFR-002 | Seam interception preserved | 100% of the ~370 patch("...agent.tasks.<sym>") call sites (~40 distinct symbols) still intercept the production call path after each relocation — evidenced per WP by the FR-002 seam checklist + interception check, not merely by "tests pass"; zero weakened or deleted assertions. | Correctness | High | Open |
| NFR-003 | Static gates | ruff and mypy --strict report zero issues on every changed src file, and strict mypy is run on changed src+test files together (Wave 1 lesson: attr-defined errors only surface with both in scope). | Maintainability | High | Open |
| NFR-004 | Whole-file ceiling | tasks.py total LOC ≤ the FR-011 enforcement ceiling (min(achieved, 1400)); measured by the gate in CI. | Maintainability | High | Open |
| NFR-005 | Targeted validation cost | Each WP declares and runs its targeted test surface — which MUST include test_tasks_cli_contract_coord.py for any WP touching commit routing or a _do_* orchestrator — not the full 26k suite; the full-suite gate is reserved for mission-level post-merge validation (charter Testing Requirements). | Process | Medium | Open |
Constraints
| ID | Title | Constraint | Category | Priority | Status |
|---|---|---|---|---|---|
| C-001 | #2300 divergence untouched | The coord+protected skip-vs-refuse divergence (move_task skips-exit-0 via the _skip_target_branch_commit pre-gate; mark_status/map_requirements refuse-exit-1 through the shared _protected_branch_status_commit_error) is preserved verbatim. Its pins are test_tasks_cli_contract_coord.py T004/T005 (incl. the wrong-leg detector) — mandatory in the targeted set of every WP that touches commit routing. Relocation must not route the three commands through a helper arrangement that adds or drops the pre-gate for any of them. | Scope | High | Open |
| C-002 | Base is merged main | The mission bases on upstream/main 381db8d5f (Wave 1 merged, including the d8a0c8c8f stabilization); all debrief line references are superseded by the squad-verified re-census in this spec. | Technical | High | Open |
| C-003 | Executable contract | This is a pure-parity refactor: the two contract harnesses + the byte-freeze suite + the new non-vacuous gates are the executable contract (ATDD-first per charter C-011 in its parity form — the byte-freeze suite is committed red-first-style BEFORE the routing change it guards, and every gate is proven red on a synthetic violation before it counts). | Process | High | Open |
| C-004 | #2173 seam discipline | No new rival seams: one production adapter per port; ports injected at the builder/shell boundary via the default-param idiom; the frozen context objects never carry live adapters. | Architectural | High | Open |
| C-005 | Terminology canon | All moved and new code, docstrings, and messages use Mission/--mission (never feature/--feature except pre-existing hidden aliases); the terminology guard (tests/architectural/test_no_legacy_terminology.py) passes before every push. | Governance | Medium | Open |
| C-006 | Gate non-vacuity | Every gate added (FR-007 AST gate, FR-011 LOC gate) ships with a concrete floor, self-mutation/synthetic-violation tests (per evasion form for FR-007), and shrink-only ratchet semantics where an allowlist/baseline exists — including the FR-009 prohibition on absorbing tasks-domain paths into _gate_coverage_baseline.json (DIRECTIVE_043); no mission-diff-scoped assertions. | Governance | High | Open |
| C-007 | Pre-existing failures reported | Any pre-existing test failure encountered is reported as a GitHub issue before being treated as baseline (charter Pre-existing Failure Reporting Rule); Wave 1's known pre-existing RED arch gates are cross-base-verified, not absorbed. | Process | Medium | Open |
Key Entities
State dataclass + port factory; the relocation unit of Stream B.
squad-identified missing piece of the per-family split.
exists to break the ports ↔ command-modules cycle risk.
(see Domain Language).
one JSON emission site.
the CI gate that selects it; the boyscout deliverable's evidence.
- Command family module: sibling module owning one family's orchestrator + glue +
- Shared helpers module: the named home for the ~30 cross-family helpers; the
- Adapters module (
tasks_command_adapters.py): owns the port-seam adapter classes; - Seam bridge: lazy parent-module attribute routing preserving patch interception
- Byte-freeze case: one recorded CLI invocation with byte-exact expected stdout for
- Marker-census artifact: the committed mapping of each tasks-domain test file to
Success Criteria (mandatory)
Measurable Outcomes
(min(achieved, 1400) LOC), enforced by a CI-visible architectural gate that fails on a synthetic violation; the recorded rationale states the delta from 4569.
directory glob, enforced by a CI-visible AST gate with a per-form self-mutation proof.
are byte-identical on the mission's final commit vs the mission base — zero behavioral diffs, with the ratchet re-pointed (FR-012), not weakened.
production call path post-relocation, evidenced by the per-WP seam checklists and interception checks; zero deleted or weakened assertions.
committed glob) selected by no CI gate; _gate_coverage_baseline.json contains no path under the glob; #2034 carries the refreshed re-census comment.
together); terminology guard green.
- SC-001:
tasks.pyis at or under the FR-011 enforcement ceiling - SC-002: 0 inline
json.dumps(all evasion forms) across the command-surface - SC-003: 43/43 contract-harness cases pass unmodified AND 13/13 byte-freeze cases
- SC-004: 100% of the ~370 patch-seam call sites (~40 symbols) intercept the
- SC-005: The committed marker-census artifact shows 0 tasks-domain files (per the
- SC-006: Zero
ruff/mypy --strictfindings on changed files (src+tests checked
Assumptions
accepted relocation template — including its actual seam mechanism (lazy parent-module attribute routing) and its no-__all__-on-siblings precedent.
avoidance); consolidation elsewhere requires a recorded no-cycle argument.
5 — the shared-helpers module and the seam-bridge/adapters foundation are their own WPs before any family move; the render-seam WP is sequenced before or with the status-family move (the _StatusRender ordering edge case); the boyscout is small.
separate mission; this mission only makes its own domain clean and refreshes the facts.
posted 2026-07-02); #2034 partial-claim comment with re-census posted 2026-07-02.
- The
mission.pydegod (kitty-specs/decompose-mission-god-module-01KVXHF8/) is the - The dedicated
tasks_command_adapters.pyis the default adapter home (import-cycle - Planning guidance (squad-adjusted sizing): expect 8–10 WPs, not the debrief's
- The repo-wide #2034 structural fix (marker gate redesign or global auto-marking) is a
- Tracker hygiene per charter: #2305 claimed (operator assigned, mission-naming comment
Campsite Folds (domain-matched, pre-plan squad 2026-07-02)
Per charter standing order 2 (fold only domain-matched debt), the pre-plan related-issues squad identified exactly three folds — everything else in #1931/#2071 is out-of-domain and stays with its epic:
base — Wave 1's inventory.md records the tasks.py sink at :1325, actual is :1326. Fold: 1-line inventory correction as a pre-step; the inventory row then moves with the move_task family relocation (FR-001/FR-002 seam checklist item).
attr-defined on _get_latest_review_cycle_verdict (re-point the import in the WP that relocates the symbol) and test_tasks.py:1028 redundant-cast (1-line removal in the first WP touching that file).
- #2306 (filed per C-007):
test_untrusted_path_containmentis RED on the mission - mypy strict inline folds (no separate issue; NFR-003 scope):
test_tasks.py:26
Parallel-work state (recorded so it isn't re-investigated): zero active collisions as of 2026-07-02 — no open upstream PR or live branch touches the tasks surface; design/degod-tasks-2116 is merged dead lineage. Latent risks fenced by comments: #2300 (do-not-start-until-Wave-2-lands warning posted), #2289 (tasks_ports.py ownership note posted), #2034 (dormant; assignee has no open PRs). Expect stale-assertion analyzer false-positive noise at every WP merge (#2031 — analyzer is intra-file only).
Non-Goals / Deferred
disposition (FR-008).
MissionResolver port — are their own missions).
deliberate and rationale-recorded, not a wart.
- #2300 — skip-vs-refuse unification (behavior change; own mission; C-001 guards it).
- Repo-wide marker-gate redesign (#2034 structural fix beyond the tasks domain).
- The unshim cluster #2289–#2293 — except the single
tasks_ports.pyshim - New ports or port extractions (#2173 Phase 1/Phase 2 work — canonicalizer gate,
- Any decision-logic change — the pure cores from Wave 1 are consumed as-is.
- Relocating
agent_tasks_ports.pyfrom its top-level placement — squad-verified as