Work Packages: Mission Type Guard Registry
Mission: mission-type-guard-registry-01KZY2FG Mission ID: 01KZY2FGYX2B90XXDD1DM3M95B Branch: kitty/mission-mission-type-guard-registry-01KZY2FG (planning base == merge target; branch-context --json confirms recommended_strategy: stay, primary_branch: main) Source of truth: spec.md · plan.md — no research.md / data-model.md / contracts/ / quickstart.md for this mission (plan.md's own Project Structure section states why: the spec's verified-figures table and the plan's Seam & Module Placement / Contracts sections already carry the design research and data-shape decisions inline). Total subtasks: 12 across 2 work packages
> This file is the WP manifest. Each WP has a prompt at tasks/WP0N-slug.md. Status lives in > status.events.jsonl, not in frontmatter. Do not hand-edit lane state — use > spec-kitty agent tasks move-task.
Scope carried forward from plan.md (binding, not re-derived here)
This mission implements exactly what spec.md's C-003 and plan.md's Charter Check bound it to: the _GUARD_TABLES registry (FR-001/FR-006), plan's own guard table plus the one-line _PRESENCE_FILE_TAGS fix (FR-002), the strict/tolerant split (FR-003–FR-005, C-001, C-002), spec-kitty doctor mission-type --json [--fail-on ...] (FR-007–FR-009), and tests for all of the above (FR-010, FR-011). Out of scope, and no WP below touches any of it: the project-wide doctrine-override hatch (_project_has_doctrine_overrides, src/charter/mission_type_profiles.py:1041), the two divergent meta readers (mission_type_profiles.py:681 vs mission.py:542), the dashboard's silent "software-dev" default (dashboard/handlers/features.py:68), the unverified wider ~22-site census, any roster/validation check inside validate_meta/write_meta (C-004), and modeling guards as DRG graph primitives. A follow-up tracking issue naming the four deferred sites is close-out work for this mission, not a WP.
No campsite-clean-first WP
Per plan.md's Campsite-Clean Scope section: the touched lines (runtime_bridge_cores.py:348-567, runtime_bridge.py:775-804 (re-verified post-#3346-rebase, shifted +105 lines from 670-699, same code), runtime_bridge_composition.py:427-486) were checked first-hand for pre-existing Sonar findings, complexity violations, or stale in-code citations, and none were found. This mission has no distinct campsite-clean-first commit — Standing Order #2's determination requirement is satisfied by stating this explicitly, not by silently skipping it. WP01's first commit is its FR-010 ATDD red-first test, not a tidy-up commit.
Baseline of record (captured in plan.md, cited not re-run)
7deadff0a4f3dfd2744b5e1e35680c0d70f4565e — 784 passed, 0 failed: tests/runtime/test_bridge_cores.py + test_bridge_composition.py (71 passed) and tests/next/ + tests/specify_cli/next/ + tests/integration/test_custom_mission_runtime_walk.py (713 passed). Verified first-hand this session: git log --oneline 7deadff0a4f3dfd2744b5e1e35680c0d70f4565e..HEAD -- src/runtime/next/ src/specify_cli/cli/commands/doctor.py src/specify_cli/cli/commands/_identity_audit.py tests/specify_cli/cli/commands/test_doctor_cli_surface_golden.py returns empty — no drift in any file this mission touches since the baseline was captured, so every line-number citation below (re-verified against the current checkout, not carried over from plan.md unchecked) is still accurate. #3284 (23 known-red main-suite failures) is a separate, pre-existing, unrelated tracking issue — not to be conflated with this baseline or fixed by either WP below.
Each WP's own targeted test surface and pytest invocation are stated in its own prompt file (charter Testing Requirements / C-005) — neither WP runs the full ~17,000-test suite.
Dependency table
| WP | Title | Requirements | Depends on | Subtasks |
|---|---|---|---|---|
| WP01 | Guard-table registry, strict/tolerant split, and plan's guard table | FR-001, FR-002, FR-003, FR-004, FR-005, FR-006, FR-010, FR-011, NFR-001, NFR-002, NFR-003, C-001, C-002 | — | T001–T007 |
| WP02 | spec-kitty doctor mission-type command | FR-007, FR-008, FR-009, NFR-004 | — | T008–T012 |
WP02 has no functional dependency on WP01 (plan.md's IC-05 note: "independent of IC-01-04 ... sequenced after them in the PR narrative since it is the 'diagnosability' half of the mission, not the 'fix' half"). Both are marked concurrently claimable below.
Write-scope disjointness (concurrent-claim check)
| WP | Owned files |
|---|---|
| WP01 | src/runtime/next/runtime_bridge_cores.py, src/runtime/next/runtime_bridge_composition.py, src/runtime/next/runtime_bridge.py, src/runtime/next/runtime_bridge_io.py, tests/runtime/test_bridge_cores.py, tests/runtime/test_bridge_composition.py, tests/runtime/test_bridge_io.py |
| WP02 | src/specify_cli/cli/commands/doctor.py, src/specify_cli/cli/commands/_mission_type_audit.py (new), tests/specify_cli/cli/commands/test_doctor_mission_type.py (new), tests/specify_cli/cli/commands/test_doctor_cli_surface_golden.py |
Zero file overlap between WP01 and WP02 — confirmed disjoint, both may be claimed and implemented in parallel lanes. This is the full production+test file manifest for the mission (6 production files + 5 test files — tests/runtime/test_bridge_io.py added per the TASKS-VERIFY-002 fix below, alongside WP01's other owned test files); no file outside this list is touched by either WP. The orchestrator has independently confirmed zero collisions against the 18 other currently-open PRs' file sets against this same manifest.
Chokepoints (shared CI gates — not optional to skip, per plan.md's Gate Set)
Neither of these serializes WP01 against WP02 at the implementation level (disjoint write scopes, no shared file), but both WPs land in one PR (see PR Shape below), so both gates below run against the PR's aggregate diff and a regression from either WP can fail the whole PR:
1. Golden CLI-surface contract test (tests/specify_cli/cli/commands/test_doctor_cli_surface_golden.py, test_registered_command_names_match_frozen_subcommands, currently pinning 19 subcommands by frozenset equality). WP02 registers a 20th (mission-type) — the golden test's FROZEN_SUBCOMMANDS / EXPECTED_OPTIONS / EXPECTED_HELP must be updated in the same commit that registers the new command (WP02/T011), or CI fails deterministically on this frozen-contract check. This is WP02-internal but load-bearing: plan.md's Contracts section calls it "the one place in this mission where 'what looks like a docs-only test file' is actually load-bearing production-contract enforcement." 2. PR diff-coverage critical-path gate (.github/workflows/ci-quality.yml, --fail-under=90 --include 'src/runtime/next/*', fed by integration-tests-next's --cov=src/runtime/next report). Every WP01 production file lives under src/runtime/next/, inside this gate's scope. WP01's ATDD commits (T001/T002) and the implementation commit (T003–T006) are expected to satisfy the 90% floor on new/changed lines, but this must be verified locally with uv run diff-cover against the base branch before push (plan.md Gate Set, PLAN-VERIFY-001's fix), not assumed.
Both gates run on every push of this mission's single PR (not only at merge time) — see plan.md's Gate Set table for the full always-on-vs-path-gated breakdown.
Requirement coverage (for map-requirements)
| WP | Requirements |
|---|---|
| WP01 | FR-001, FR-002, FR-003, FR-004, FR-005, FR-006, FR-010, FR-011, NFR-001, NFR-002, NFR-003, C-001, C-002 |
| WP02 | FR-007, FR-008, FR-009, NFR-004 |
C-003, C-004, C-005 are mission-scope/process constraints (bounded scope, no validate_meta roster check, targeted test surface) satisfied by this tasks.md's own structure and each WP's test-strategy section, not by a specific code change inside either WP — they are not mapped to a WP via map-requirements for that reason (consistent with how constraints of that shape are generally not code-mapped elsewhere in this repo's missions).
PR Shape — carried forward from plan.md, re-verified against this WP slicing
One PR for the whole mission, per plan.md's own decision. Re-verified here against the actual 2-WP slice, not just repeated: the full touched-file set is 6 production files + 5 test files (see Write-scope disjointness above), the behavior change is one conceptual unit (close one silent-fallback defect class across two call paths, plus one diagnostic command reading the same underlying facts), and splitting into 2 WPs did not surface any migration-chain touch, contract move, or cross-repo coordination trigger that plan.md's own split-it checklist would flag. Both WPs are small enough to review in one sitting individually. Sizing the combined diff precisely rather than folding it into one undifferentiated figure (TASKS-DECOMP-001/002 fix): WP01's Commit 3 (T003-T006) is two different sizes in two different places — T004-T006 are genuinely small, ~10 lines total across 3 files (runtime_bridge_io.py, runtime_bridge_composition.py, runtime_bridge.py), each a one-call-site change; T003 alone, in runtime_bridge_cores.py, adds a new _GUARD_TABLES dict, a new UnregisteredMissionFamilyError exception class (carrying a mandated cross-reference docstring — this codebase's own analogous exception, charter.mission_type_profiles.UnknownMissionTypeError, runs 13 lines, a concrete size anchor), a new evaluate_guards_strict function, the new ~5-way _evaluate_plan_guards function, and a rewritten evaluate_guards — five new/changed symbols in one file, realistically 40-60 new/changed lines, not "~10 lines." WP02's _mission_type_audit.py is sized per WP02's own Context & Constraints guidance rather than as an independent target: it combines the domain-classifier and CLI-glue roles of two precedent modules (src/specify_cli/status/identity_audit.py, 361 lines; src/specify_cli/cli/commands/_identity_audit.py, 346 lines — 707 lines combined, wc -l verified), and WP02's own prompt explicitly instructs the implementer not to treat either single precedent's line count as a target — so the new sibling module should be expected to land at the two precedents' combined LOC order of magnitude, roughly 600-700 lines, not the smaller "~300-line" figure this section previously stated. With both figures corrected (WP01: ~10 lines in 3 files, plus ~40-60 lines / 5 symbols in one file; WP02: one new CLI command + one new ~600-700-line sibling module), the combined diff is still small enough to review in one sitting as a whole — a review budget of one sitting for a handful of new symbols in one existing runtime file, three tiny call-site edits, and one new sibling module of that scale is realistic for this codebase's typical review cadence. This tasks author's conclusion: one PR remains correct. If a reviewer disagrees after seeing the actual diff size once WP02's _mission_type_audit.py is drafted (plan.md's own PR Shape section already flags this exact risk — "if the golden-contract update turns out to ripple into other doctor-surface tests not identified here" — as a stop-and-rescope signal), that is a decision for the orchestrator/operator to make, not something this tasks.md pre-decides.
Work Package Sections
WP01 — Guard-table registry, strict/tolerant split, and plan's guard table
NFR-001, NFR-002, NFR-003, C-001, C-002
registry; split the shared evaluate_guards call into a strict lookup (evaluate_guards_strict, raises UnregisteredMissionFamilyError) that _check_cli_guards calls directly and a tolerant, WARNING-logging wrapper that _check_composed_action_guard calls directly; author _evaluate_plan_guards and register it under "plan"; fix the _PRESENCE_FILE_TAGS gap for research.md. This is IC-01 through IC-04 from plan.md's Implementation Concern Map, landed as one coherent, single-concern change (plan.md's own determination: these four concerns must land together, not be further split by an artificial tidy-up pre-commit).
- Prompt:
tasks/WP01-guard-table-registry-and-plan-guards.md - Depends on: —
- Requirement refs: FR-001, FR-002, FR-003, FR-004, FR-005, FR-006, FR-010, FR-011,
- Goal: Replace
evaluate_guards's if/if/fall-through with an explicit_GUARD_TABLES - Included subtasks (tracked via
spec-kitty agent tasks mark-status): T001–T007
WP02 — spec-kitty doctor mission-type command
on doctor identity, classifying every mission under kitty-specs/ into the FR-008 6-state taxonomy. This is IC-05 + IC-06 from plan.md's Implementation Concern Map.
- Prompt:
tasks/WP02-doctor-mission-type-command.md - Depends on: —
- Requirement refs: FR-007, FR-008, FR-009, NFR-004
- Goal: Ship
spec-kitty doctor mission-type --json [--fail-on <states>], modeled directly - Included subtasks (tracked via
spec-kitty agent tasks mark-status): T008–T012
Subtask Index
| ID | Description | WP | RED/impl | Parallel? |
|---|---|---|---|---|
| T001 | RED: FR-010 ATDD pin — plan/review target-shape assertion ([]), genuinely RED against base; plus the FR-002/User-Story-2-AC3 plan/research tightening pin, the plan/specify branch pin (User Story 2 AC2) plus plan/plan and fail-closed-else branch pins (hardening beyond spec.md's literal Acceptance Scenarios), and a disk-backed gather_artifact_presence pin for the research.md presence tag | WP01 | RED | No |
| T002 | RED: FR-011 ATDD pin — unregistered-family fall-through, 3 assertions (evaluate_guards_strict raises; _check_cli_guards propagates via injection seam; _check_composed_action_guard returns [] + WARNING log) | WP01 | RED | No |
| T003 | Add _GUARD_TABLES, UnregisteredMissionFamilyError, evaluate_guards_strict, tolerant evaluate_guards wrapper, _evaluate_plan_guards in runtime_bridge_cores.py | WP01 | impl | No |
| T004 | Add "research.md" to _PRESENCE_FILE_TAGS in runtime_bridge_io.py; update module docstring "three"→"four" | WP01 | impl | Yes (with T005/T006) |
| T005 | _check_composed_action_guard in runtime_bridge_composition.py: call evaluate_guards_strict directly, catch, WARNING-log, return [] | WP01 | impl | Yes (with T004/T006) |
| T006 | _check_cli_guards in runtime_bridge.py: call evaluate_guards_strict directly, no catch | WP01 | impl | Yes (with T004/T005) |
| T007 | Verify T001/T002 GREEN; zero new reds vs. 784-baseline; local diff-cover check against src/runtime/next/* | WP01 | impl | No |
| T008 | RED: FR-008/SC-005/SC-006 ATDD pin — test_doctor_mission_type.py, fixture tree with one mission per taxonomy state + FR-008 boundary case + --fail-on behavior + an automated NFR-004 timing regression test (synthetic 200-mission fixture, elapsed < 2.0) | WP02 | RED | No |
| T009 | Implement _mission_type_audit.py — MissionTypeState, classify_mission_type, audit_mission_types, summarize_mission_types, run_mission_type_audit | WP02 | impl | No |
| T010 | Implement doctor.py mission-type thin @app.command shell | WP02 | impl | Yes (with T011) |
| T011 | Update golden contract test: FROZEN_SUBCOMMANDS, EXPECTED_OPTIONS["mission-type"], EXPECTED_HELP["mission-type"], 19→20 count comment | WP02 | impl | Yes (with T010) |
| T012 | Verify T008 GREEN, including the automated NFR-004 timing test; manual timing spot-check as a belt-and-suspenders confirmation; targeted test surface green | WP02 | impl | No |
> Replace-placeholder note: this file was hand-authored per the mission's tasks-phase brief > (no /spec-kitty.tasks LLM step available in this dispatch — the canonical tasks-template.md > structure is followed directly). spec-kitty agent tasks map-requirements and > spec-kitty agent mission finalize-tasks are the canonical tools that validate and finalize > this file's dependency declarations and write lanes.json — run after the WP prompt files > exist.