Tasks: Charter as Sole Door: Close Bypass Access Paths
Input: plan.md, spec.md, research.md, data-model.md, contracts/, quickstart.md from kitty-specs/charter-sole-door-bypass-closure-01KZ3WAA/
Subtask completion is event-sourced (spec-kitty agent tasks mark-status <Txxx> --status done), not a markdown checkbox. Rows below are reference only.
Restructured by a post-tasks adversarial squad (architect-alphonso, reviewer-renata, debugger-debbie, paula-patterns) after first generation — see each WP's Activity Log / History for what changed and why. Headline fixes: (1) CRITICAL — WP02/WP03/WP04's dependencies: [] frontmatter never actually declared WP01, despite the prose saying so everywhere (found independently by 3 of 4 delegates); (2) former WP07 merged into WP01 (both edited src/charter/resolver.py; the split forced an awkward 3-lane serialization for no benefit); (3) the durability-gate WP (former WP09) had 2 of its 5 gates moved to the WPs whose surface they actually guard (WP04 absorbed Gate 5, WP06 absorbed Gate 4); (4) several accessor method names, line-number citations, and one WP's core premise were corrected after independent verification against the live code.
Subtask Index
| ID | Description | WP | Parallel |
|---|---|---|---|
| T001 | Implement the pinned agent_profile_repository accessor on charter.resolver.DoctrineService | WP01 | |
| T002 | Unify the two named builder functions into one (active_languages always computed, org_roots always self-resolved) | WP01 | |
| T003 | Retarget org_layer.py:244,275 and generate.py:56 onto the unified builder; fix fail-open except ImportError: pass | WP01 | |
| T004 | Regression test: unified builder identical output across ALL 9 gated properties (one pass, not staged) | WP01 | |
| T005 | ATDD test: accessor semantics (register_overlay(), get_provenance()) | WP01 | |
| T026-T031 | Add 6 mechanical properties (directives/tactics/styleguides/toolguides/mission_step_contracts/glossary_packs) | WP01 | [P] |
| T032 | Bare-project equality regression test per kind (6 tests): wrapped == unwrapped inner service | WP01 | |
| T006 | Capture pre-mission p95 baseline (raw timing series) for spec-kitty agent tasks status on a 100+-WP fixture | WP02 | |
| T007 | Migrate runtime_bridge_io.py:576 onto the factory via WP01's accessor (resolve_profile) | WP02 | [P] |
| T008 | Migrate projection.py:84 onto the factory via WP01's accessor (register_overlay only) | WP02 | [P] |
| T009 | Migrate tasks_status_cmd.py:712,823; re-measure p95 in the same session, compare against T006 | WP02 | |
| T010 | Confirm charter/profile_resolution.py:81 is a genuine bootstrap case, document, do NOT migrate | WP02 | [P] |
| T011 | Migrate charter/compiler.py:802 onto the factory | WP03 | [P] |
| T012 | Migrate _doctrine_asset.py:75 onto the factory | WP03 | [P] |
| T013 | Migrate _doctrine_collect.py's 4 sites (lines 193/283/420/828) onto the factory's unfiltered mode | WP03 | |
| T014 | Regression test: unfiltered mode returns catalog identical to the raw unwrapped service | WP03 | |
| T015 | Migrate registry.py:64 off ._inner.agent_profiles onto WP01's accessor (get_provenance) | WP04 | [P] |
| T016 | Migrate org_profiles.py:117 off ._inner.agent_profiles onto WP01's accessor (get_provenance) | WP04 | [P] |
| T017 | Gate 5 (absorbed): mission-wide zero-tolerance ._inner-on-doctrine-service gate, self-mutation proven, true-negative tested | WP04 | |
| T018 | Add resolution methods (distinct names, no collision) to charter.resolver.DoctrineService | WP05 | |
| T019 | Resolve the construction-contract mismatch for CharterTemplateResolver's one real caller (repo_root vs missions_root) | WP05 | |
| T020 | Retarget CharterTemplateResolver (thin delegate or retire in favour of direct factory use) | WP05 | |
| T021 | Regression test: tier resolution via factory matches the old direct-call result | WP05 | |
| T022 | Make builtin_missions_root() a thin delegate to MissionTemplateRepository.default_missions_root() | WP06 | [P] |
| T023 | Retarget runtime/home.py's dev_roots fallback onto default_missions_root() | WP06 | [P] |
| T024 | Equality regression test for both retargeted sites | WP06 | |
| T025 | Document home.py's residual #2986-shared risk, the #3091-deferred convergence, AND 3 newly-found untouched residual hardcodes | WP06 | |
| T040 | Gate 4 (absorbed): hardcoded missions-root path-literal gate + self-mutation proof | WP06 | |
| T034 | Add activation filtering to resolve_mission_type_context() (NOT MissionTypeProfileRepository's own file) | WP08 | |
| T035 | Bare-project regression test: set-equality against builtin_mission_type_id_set() | WP08 | |
| T036 | Regression test: a subset of activated mission-types returns only that subset | WP08 | |
| T037 | Gate 1: zero-tolerance, qualname-resolving AgentProfileRepository gate + composite-key exclusions + self-mutation proof | WP09 | [P] |
| T038 | Gate 2: zero-tolerance, qualname-resolving DoctrineService gate + composite-key exclusions + self-mutation proof | WP09 | [P] |
| T039 | Gate 3: doctrine.resolver direct-import forward-looking regression guard + self-mutation proof | WP09 | [P] |
| T042 | Add issue-matrix.json rows + citable docs record for #2986/#3036/#3039/#3091/#3022 | WP10 | [P] |
| T043 | Post a tracker comment on each of the 5 deferred issues; paste gh issue view --comments output as evidence | WP10 | [P] |
| T044 | Add the NFR-004 CHANGELOG.md entry (9-kind gating, .kittify/profiles exclusion, deferred pack_paths convergence) | WP10 | [P] |
Work Packages
WP01 — Extend the charter factory: accessor, builder unification, 6 mechanical kinds (Priority: P1)
Goal: One canonical builder; one new lineage/mutation accessor; 6 new activation-gated properties. Requirements: C-001, FR-001, FR-005, FR-008, FR-010. Independent test: construct the unified builder from both former call sites' inputs — identical output across all 9 gated properties; bare-project equality holds for each of the 6 new kinds. Subtasks: T001-T005, T026-T032 (10 subtasks — over the 3-7 ideal / at the 10 max, a deliberate consolidation the post-tasks squad recommended over a 3-way split that all touched the same file). Dependencies: none (first). Estimated size: ~240 lines (still within the 700-line hard cap despite subtask count, since most of T026-T031 is copy-paste of one pattern). Risks: the accessor's semantics are pinned in the prompt — implement against that text, do not re-derive. Write T004's identical-output proof once, against the full 9-property surface, not staged.
WP02 — Migrate AgentProfileRepository sites + NFR-005 measurement (Priority: P1)
Goal: Close 4 of 5 in-scope direct AgentProfileRepository construction sites (the 5th, profile_resolution.py:81, is confirmed a genuine bootstrap case and is documented, not migrated); prove no perf regression. Requirements: FR-001, NFR-005. Independent test: grep confirms zero direct construction at the 4 migrated sites; p95 latency (raw timing series, same session) within 10% of baseline. Subtasks: T006-T010. Dependencies: WP01 (fixed — the original frontmatter omitted this; found by 3 of 4 post-tasks squad delegates independently). Estimated size: ~400 lines. Risks: T006 MUST run before T009 touches tasks_status_cmd.py. projection.py:84 needs register_overlay() only (not get_ancestors(), which is unused there).
WP03 — Migrate raw DoctrineService sites (Priority: P1)
Goal: Close the 6 originally-flagged raw DoctrineService construction sites (the 3 additional sites found by the post-plan squad — org_layer.py/generate.py — are WP01's). Requirements: FR-002. Independent test: grep confirms zero raw construction outside the factory/builder; _doctrine_collect.py diagnostics still see the full unfiltered catalog. Subtasks: T011-T014. Dependencies: WP01 (fixed — see WP02's note). Estimated size: ~350 lines. Risks: a plain activation-aware swap on _doctrine_collect.py silently narrows doctor/health output — must use pack_context=None explicitly. Line numbers are 193/283/420/828 (corrected — the original citations had drifted +2 lines).
WP04 — Close the ._inner reach-around + Gate 5 (Priority: P1)
Goal: Eliminate the escape hatch that would otherwise defeat every other WP's gating; ship its own durability gate (absorbed from the former WP09). Requirements: FR-010, FR-007. Independent test: zero ._inner attribute access on a doctrine-service-typed expression outside src/charter/, gate-enforced with a self-mutation proof (both true-positive and true-negative tested). Subtasks: T015-T017. Dependencies: WP01 (fixed). Estimated size: ~210 lines. Risks**: both sites need get_provenance() (corrected — not register_overlay()). The gate must not false-positive on unrelated ._inner attributes in auth/transport.py/events/decision_log.py.
WP05 — Consolidate the template/command resolver axis onto the factory (Priority: P1)
Goal: Fold CharterTemplateResolver (a second charter-layer door) into the factory — a reframed goal (post-tasks squad correction: the original "stops importing doctrine.resolver directly" premise was false; nothing outside src/charter/ ever imported it). Requirements: FR-003. Independent test: tier resolution via the factory matches the old CharterTemplateResolver result; doctrine/resolver.py itself is untouched. Subtasks: T018-T021. Dependencies: WP01 (T018 adds methods to src/charter/resolver.py, WP01's sole declared owner — a sequenced, dependency-gated out-of-map edit). Estimated size: ~340 lines. Risks**: pick new method names — do NOT reuse CharterTemplateResolver's existing resolve_command_template/resolve_content_template names, which have different signatures (a real naming-collision risk found by the squad).
WP06 — Consolidate missions-root path hardcodes + Gate 4 (Priority: P1)
Goal: Retarget the 2 duplicate missions-root hardcodes onto one promoted authority; ship its own durability gate (absorbed from the former WP09); name 3 newly-found untouched residual duplicates honestly. Requirements: FR-004, FR-007. Independent test: equality regression test — both retargeted sites resolve identically to default_missions_root(); gate self-mutation proof fails on a reintroduced hardcode. Subtasks: T022-T025, T040. Dependencies: none (parallel). Estimated size: ~300 lines. Risks: do not claim pack_paths.built_in_dir convergence. Do not silently omit the 3 newly-found residual hardcodes (kernel/paths.py, template/manager.py, list_cmd.py — verify exact citations before writing, the squad's line numbers were not independently re-checked) — name them, don't fix or hide them.
WP08 — Mission-type activation gating (Priority: P1)
Goal: Gate the mission-type token via resolve_mission_type_context(), not DoctrineService or MissionTypeProfileRepository's own file (ownership boundary with WP06). Requirements: FR-006. Independent test: a bare project's mission-type set equals builtin_mission_type_id_set(); an activated subset returns only that subset. Subtasks: T034-T036. Dependencies: none (parallel). Estimated size: ~250 lines. Risks: T035's assertion MUST be set-equality against builtin_mission_type_id_set(), not a fakeable subset check.
WP09 — Durability gates 1-3 (Priority: P2)
Goal: Ship non-vacuous, zero-tolerance, qualname-resolving architectural gates for the three genuinely cross-cutting bypass categories (Gates 4 and 5 moved to WP06/WP04 — each only ever guarded its own WP). Requirements: FR-007, NFR-001, NFR-003. Independent test: each gate's self-mutation proof (function-local/nested injection) fails when the guarded shape is reintroduced, naming the offending site. Subtasks: T037-T039. Dependencies: WP01, WP02, WP03, WP05 (must run after these land — a gate written before the bypasses close is either vacuous or immediately red). Estimated size: ~300 lines. Risks: Gate 2's exclusion line numbers must be re-verified against the post-WP03 worktree state, not assumed. Gate 3 is a forward-looking regression guard, not proof of a WP05 closure — there was no violation outside src/charter/** to close in the first place.
WP10 — Deferred-issue tracker hygiene + CHANGELOG entry (Priority: P2)
Goal: Give the five confirmed-deferred issues a durable tracker trace, not just PR-description prose; add the CHANGELOG entry NFR-004 requires (added by /spec-kitty.analyze finding E1 — the original task breakdown had zero coverage for this NFR). Requirements: FR-011, NFR-004. Independent test: each of #2986/#3036/#3039/#3091/#3022 carries an issue-matrix.json row, a citable docs record, and a tracker comment naming this mission — with the comment's existence proven via pasted gh issue view --comments output (non-fakeable evidence requirement, post-tasks squad correction); CHANGELOG.md names all three NFR-004 items. Subtasks: T042-T044. Dependencies: none (fully independent of code; can run any time). Estimated size: ~150 lines.
Sequencing Summary
WP01 ─┬─> WP02 ──┐
├─> WP03 ──┤
├─> WP04 ──┤──> WP09 (Gates 1-3, last; depends on WP01/02/03/05)
└─> WP05 ──┘
WP06 (independent, ships its own Gate 4)
WP08 (independent)
WP10 (fully independent, any time)src/charter/resolver.py is edited by WP01 (sole declared owner: accessor, builder unification, 6 mechanical properties) and WP05 (tier-axis methods, a sequenced, dependency-gated out-of-map edit — WP05 depends on WP01, never parallel with it). The former 3-way split (WP01/WP05/WP07) was consolidated to WP01+WP05 after a post-tasks squad found the split added a serialization stage for no benefit.
MVP scope: WP01-04 (the direct-construction closure + accessor + Gate 5) delivers the largest single chunk of the "sole door" claim independently of WP05/06/08's axis/path/mission-type work.