Implementation Plan: Excise Doctrine Curation and Inline References

Mission ID: 01KP54J6W03W8B05F3P2RDPS8S (mid8: 01KP54J6) Mission slug: excise-doctrine-curation-and-inline-references-01KP54J6 Branch: main (planning base = merge target = main) Date: 2026-04-14 Spec: spec.md Tracking issue: Priivacy-ai/spec-kitty#463

Summary

Execute Phase 1 of EPIC #461 by excising every surviving carrier of the pre-DRG doctrine model on main: the curation pipeline (package, CLI, validator, _proposed/ trees), the inline tactic_refs / paradigm_refs / applies_to fields on shipped artifacts + schemas + models, the legacy transitive reference resolver, the include_proposed catalog flag, and the parallel build_charter_context / build_context_v2 split. The end state is a single context builder (renamed to build_charter_context), validators that actively reject inline refs, and src/doctrine/drg/query.py exposing a new resolve_transitive_refs() that serves the pre-existing callers (resolve_governance(), compile_charter()) without requiring a deeper Phase 3 rewrite.

The technical approach is a strict three-WP excision (WP1.1 → WP1.2 → WP1.3) guarded by the #393 occurrence-classification pattern. Every WP ships its own occurrence artifact under kitty-specs/excise-doctrine-curation-and-inline-references-01KP54J6/occurrences/WP1.<n>.yaml, and a mission-level index aggregates the "must-be-zero" set across WPs. Verification is by completeness (scripted check of the occurrence artifact against the live tree), not by semantic-diff review.

Technical Context

Language/Version: Python 3.11+ (existing spec-kitty requirement) Primary Dependencies: typer, rich, ruamel.yaml, pytest, mypy --strict (unchanged) Storage: Filesystem only — YAML doctrine artifacts, src/doctrine/graph.yaml, project-local overlays Testing: pytest with ≥90% coverage on changed code per charter; mypy --strict must pass Target Platform: The spec-kitty-cli Python package distributed via PyPI Project Type: Single project (this is pure refactor inside src/ and tests/) Performance Goals: Do not regress pytest runtime more than 5% (NFR-001); no changes to user-facing runtime performance expected since build_context_v2() is already parity-safe Constraints: No fallback, no backwards compatibility, no deprecation shim (C-001); strict #393 guardrail usage (C-002); only SOURCE templates editable (C-005); m_3_1_1_charter_rename.py migration is carved out (C-006); strict WP sequencing (C-007) Scale/Scope: ~2500 LOC deletion plus a targeted rewrite of resolve_references_transitively into a DRG-backed resolve_transitive_refs (~150 LOC). ~27 test files impacted. Twin charter packages (src/charter/ and src/specify_cli/charter/) edited in lockstep.

Charter Check

GATE: Must pass before Phase 0 research. Re-check after Phase 1 design.

Charter file: /Users/robert/spec-kitty-dev/spec-kitty-charter-14-April/spec-kitty/.kittify/charter/charter.md exists.

Governance file: /Users/robert/spec-kitty-dev/spec-kitty-charter-14-April/spec-kitty/.kittify/charter/governance.yaml does not exist on main today (the bootstrap spec-kitty charter context emits governance.yaml not found. Run 'spec-kitty charter sync'). This is an informational note, not a gate violation for this mission — Phase 1 does not depend on governance.yaml and does not write one.

Charter-derived policy summary applied to this mission:

PolicyApplies to this mission?How we honor it
typer, rich, ruamel.yaml, pytest, mypy --strictYesNo new runtime deps introduced; validator rejection shape uses existing ruamel.yaml load path
≥90% test coverage on new codeYesresolve_transitive_refs, validator rejection paths, and rewritten context builder tests must meet this bar
mypy --strict passesYesEvery WP PR blocks on mypy-strict in CI
Integration tests for CLI commandsYesWP1.1 includes an integration test that asserts spec-kitty doctrine curate fails with unknown-command; WP1.3 includes an integration test of spec-kitty charter generate against the DRG-backed compile_charter path

DIRECTIVE_003 (Decision Documentation): Every plan-phase decision is recorded in this plan.md and in the mission-level occurrence index. No decisions are deferred to implementation silently.

DIRECTIVE_010 (Specification Fidelity): Any deviation from spec.md during implementation must be logged in spec.md as an amendment and re-reviewed before merge. The resolve_transitive_refs extraction and the rename-only-at-WP1.3 sequencing are plan-phase refinements that this plan records explicitly.

Gate status: PASS. No violations. Governance-file absence is informational.

Project Structure

Documentation (this feature)

kitty-specs/excise-doctrine-curation-and-inline-references-01KP54J6/
├── spec.md
├── plan.md
├── research.md
├── data-model.md
├── quickstart.md
├── contracts/
│   ├── occurrence-artifact.schema.yaml
│   ├── validator-rejection-error.schema.json
│   ├── resolve-transitive-refs.contract.md
│   └── removed-cli-surface.md
├── occurrences/                       # created by /spec-kitty.tasks + implementation
│   ├── index.yaml                     # mission-level aggregate
│   ├── WP1.1.yaml
│   ├── WP1.2.yaml
│   └── WP1.3.yaml
├── checklists/
│   └── requirements.md
├── meta.json
└── tasks.md                           # /spec-kitty.tasks output

Source code (repository root)

src/
├── charter/                           # library package — edited in WP1.2 + WP1.3
│   ├── __init__.py                    # export surface flipped in WP1.3
│   ├── catalog.py                     # drop include_proposed in WP1.3
│   ├── compiler.py                    # swap resolver import in WP1.3
│   ├── context.py                     # collapse to one builder in WP1.3
│   ├── generator.py                   # untouched (re-export layer)
│   ├── reference_resolver.py          # DELETE in WP1.3
│   ├── resolver.py                    # swap resolver import in WP1.3
│   ├── schemas.py                     # strip applies_to from Directive in WP1.2
│   └── sync.py                        # untouched
├── specify_cli/
│   ├── charter/                       # CLI-facing twin of src/charter/ — lockstep edits
│   │   ├── __init__.py
│   │   ├── compiler.py
│   │   ├── context.py
│   │   └── resolver.py
│   ├── cli/commands/
│   │   ├── __init__.py                # remove doctrine Typer registration in WP1.1
│   │   ├── charter.py                 # call-site flip in WP1.3
│   │   ├── doctrine.py                # DELETE in WP1.1
│   │   └── agent/workflow.py          # call-site flip in WP1.3
│   ├── missions/software-dev/
│   │   └── command-templates/         # SOURCE templates — edited in any WP that changes prose
│   ├── next/
│   │   └── prompt_builder.py          # call-site flip in WP1.3
│   ├── runtime/
│   │   └── doctor.py                  # resolver call-site touched in WP1.3
│   └── validators/
│       └── doctrine_curation.py       # DELETE in WP1.1
├── doctrine/
│   ├── agent_profiles/validation.py   # reject inline refs (if any) in WP1.3
│   ├── directives/
│   │   ├── _proposed/                 # DELETE entire dir in WP1.1
│   │   ├── models.py                  # strip inline fields in WP1.2
│   │   ├── validation.py              # reject inline refs in WP1.3
│   │   └── *.directive.yaml           # strip tactic_refs in WP1.2 (8 files)
│   ├── paradigms/
│   │   ├── _proposed/                 # DELETE entire dir in WP1.1
│   │   ├── models.py                  # strip inline fields in WP1.2
│   │   ├── validation.py              # reject inline refs in WP1.3
│   │   └── *.paradigm.yaml            # strip tactic_refs in WP1.2 (3 files)
│   ├── procedures/
│   │   ├── _proposed/                 # DELETE entire dir in WP1.1
│   │   ├── models.py                  # strip inline fields in WP1.2
│   │   ├── validation.py              # reject inline refs in WP1.3
│   │   └── *.procedure.yaml           # strip tactic_refs in WP1.2 (2 files)
│   ├── styleguides/_proposed/         # DELETE (empty) in WP1.1
│   ├── tactics/
│   │   ├── _proposed/                 # DELETE entire dir in WP1.1
│   │   └── validation.py              # reject inline refs in WP1.3
│   ├── toolguides/
│   │   ├── _proposed/                 # DELETE (empty) in WP1.1
│   │   └── validation.py              # reject inline refs in WP1.3
│   ├── curation/                      # DELETE entire package in WP1.1
│   ├── drg/
│   │   ├── loader.py                  # untouched
│   │   ├── models.py                  # untouched
│   │   ├── query.py                   # ADD resolve_transitive_refs() in WP1.3
│   │   └── validator.py               # untouched
│   ├── graph.yaml                     # additive-only remediation in WP1.2 if missing edges found
│   └── schemas/
│       ├── directive.schema.yaml      # strip inline ref fields in WP1.2
│       ├── paradigm.schema.yaml       # strip inline ref fields in WP1.2
│       └── procedure.schema.yaml      # strip inline ref fields in WP1.2

tests/
├── charter/
│   ├── test_context.py                # REWRITE in WP1.3 to target single builder
│   ├── test_context_parity.py         # DELETE in WP1.3 (after coverage subsumed)
│   └── test_reference_resolver.py     # DELETE in WP1.3 (replaced by drg/query tests)
├── doctrine/
│   ├── curation/                      # DELETE entire tree in WP1.1
│   ├── test_inline_ref_rejection.py   # NEW in WP1.3 — one fixture per kind
│   └── drg/
│       └── test_resolve_transitive_refs.py  # NEW in WP1.3
├── cross_cutting/
│   └── test_doctrine_curation_unit.py # DELETE in WP1.1
├── agent/test_workflow_charter_context.py  # UPDATE in WP1.3
└── (various model/consistency tests)  # UPDATE inline field assertions in WP1.2

scripts/
└── verify_occurrences.py              # NEW in WP1.1 (reused by WP1.2 / WP1.3)

kitty-specs/excise-doctrine-curation-and-inline-references-01KP54J6/occurrences/
├── index.yaml
├── WP1.1.yaml
├── WP1.2.yaml
└── WP1.3.yaml

Structure Decision: Single-project Python layout. All changes land in the existing src/ and tests/ trees. No new top-level directories introduced except kitty-specs/.../occurrences/ (mission-scoped) and scripts/verify_occurrences.py (one helper script co-located with the existing scripts/ directory if present, otherwise created). The twin charter packages (src/charter/ and src/specify_cli/charter/) stay as two separate package trees — deduplicating them is explicitly out of scope for this mission.

Plan-Phase Decisions (locked)

D-1 — Context builder naming

Decision: Rename build_context_v2 → build_charter_context during WP1.3 only at the point the legacy implementation is removed. Do not front-load the rename before the live-path cutover is complete.

Rationale: Front-loading the rename would leave a window where the legacy implementation and the new (renamed) implementation coexist under the same name — a debugging nightmare. Performing the rename inside the single WP1.3 PR ensures the cutover is atomic.

Sequencing inside WP1.3: 1. Add DRG-backed resolve_transitive_refs in src/doctrine/drg/query.py. 2. Flip charter/resolver.py and charter/compiler.py (and their specify_cli/charter/ twins) to import the new function. Keep the legacy reference_resolver.py present but now unused by production paths; tests still exercise it. 3. Flip the five+ live build_charter_context() call sites to call build_context_v2() directly. 4. Run the full pytest suite at this point: legacy builder is imported by nothing in src; still exercised only by tests/charter/test_context.py (legacy) and tests/charter/test_context_parity.py. 5. Rename build_context_v2 → build_charter_context in src/charter/context.py; delete the old build_charter_context implementation; update charter/__init__.py and specify_cli/charter/__init__.py re-exports; flip the five+ call sites back to the renamed import. 6. Delete src/charter/reference_resolver.py after the last importer is moved. 7. Delete tests/charter/test_context_parity.py and tests/charter/test_reference_resolver.py only after the replacement suites (tests/charter/test_context.py rewritten + tests/doctrine/drg/test_resolve_transitive_refs.py new) land green and demonstrably cover their behavioral surface.

D-2 — Reference-resolver replacement location (Q1 = A) — amended 2026-04-14

Decision: Place the DRG-backed resolve_transitive_refs() in src/doctrine/drg/query.py alongside the existing resolve_context() primitive. Implement as a thin bucketing wrapper over the existing walk_edges(graph, start_urns, relations, max_depth) primitive. Return a ResolveTransitiveRefsResult with per-kind bucketed bare IDs (URN prefix stripped) — same field shape as legacy ResolvedReferenceGraph, plus paradigms and agent_profiles buckets that legacy lacked. See contracts/resolve-transitive-refs.contract.md for the full corrected contract.

Input type: DRGGraph (the live doctrine.drg.models.DRGGraph), NOT a fabricated MergedGraph. merge_layers() returns DRGGraph.

Relation selection: callers pass {Relation.REQUIRES, Relation.SUGGESTS} for legacy parity — these are the two relation kinds the Phase 0 migration extractor used when translating inline tactic_refs/paradigm_refs into DRG edges. Confirmed by reading doctrine.drg.query.resolve_context() (the Phase 0 reference implementation), which uses REQUIRES (transitive) and SUGGESTS (depth-limited) for artifact reachability.

Caller-side helper: introduce src/charter/_drg_helpers.py :: _load_validated_graph(repo_root) (+ src/specify_cli/charter/_drg_helpers.py twin) to encapsulate the load_graph(doctrine_root/'graph.yaml') + merge_layers(shipped, project_overlay) + assert_valid(merged) sequence. resolver.py and compiler.py each import the helper once. This avoids duplicating the 4-line graph-load sequence at every call site.

Rationale: Minimizes blast radius on the live path and keeps charter/resolver.py / charter/compiler.py / charter generate behavior-stable without requiring the deeper Phase 3 rewrite. The function is a DRG-layer capability (walk the merged graph), not a charter-layer capability, so it belongs in src/doctrine/drg/. Cycle detection is NOT relocated to this function — assert_valid() already rejects requires cycles at graph load; walk_edges is BFS-with-visited-set so suggests cycles are benign.

D-3 — Twin-package lockstep

Decision: Every edit that touches src/charter/{context,compiler,resolver,catalog,__init__}.py also touches the equivalent src/specify_cli/charter/* file in the same WP PR. Occurrence artifacts list both file paths as separate import_path entries so the verifier catches a one-sided edit.

Rationale: The two packages are a known pre-existing wart. Deduplicating them is Phase 3+ work. Phase 1 cannot introduce drift between them.

D-4 — Test architecture

Decision: Coverage is collapsed, not lost.

Rules enforced in plan:

  • WP1.1: delete curation tests (tests/doctrine/curation/**, tests/cross_cutting/test_doctrine_curation_unit.py). No coverage loss — those tests only cover deleted code.
  • WP1.2: update inline-field assertions in artifact model tests (7+ files) so they now assert the fields are absent from shipped YAMLs. No deletions.
  • WP1.3: rewrite tests/charter/test_context.py as a single-builder suite; add tests/doctrine/test_inline_ref_rejection.py (one negative fixture per kind); add tests/doctrine/drg/test_resolve_transitive_refs.py. Run the rewritten + new tests in parallel with test_context_parity.py and test_reference_resolver.py and verify green. Only then delete test_context_parity.py and test_reference_resolver.py. This is the "collapse coverage, do not create a regression hole" rule.

D-5 — Occurrence-classification artifact shape (per #393 + FR-015)

Decision: Per-WP YAML artifacts at kitty-specs/.../occurrences/WP1.<n>.yaml + mission-level aggregate at kitty-specs/.../occurrences/index.yaml. Schema defined in contracts/occurrence-artifact.schema.yaml. Verifier script scripts/verify_occurrences.py (added in WP1.1, reused by WP1.2/WP1.3) reads the artifact and asserts the "to-change" set is empty on disk.

CI enforcement: The verifier is run in each WP PR as a non-skippable check but implemented cheaply — it's a pure Python walk, no external services. If the implementation lands heavier than "cheap," per the user's guidance downgrade to a local pre-merge script + a mission-level aggregate check in the final WP PR.

D-6 — Charter Check note

Decision: Informational only, not a gate. Governance-file absence does not block Phase 1.

D-7 — Strict sequencing

Decision: WP1.1 → WP1.2 → WP1.3, each its own merged PR, each preflight-verified, each verifier-green before the next begins.

D-8 — Research scope

Decision: Three narrow Phase 0 investigations captured in research.md:

  • R-1: exhaustive inline-field inventory vs graph.yaml edge coverage — produce the "missing edges" remediation list for WP1.2.
  • R-2: behavioral-equivalence proof for resolve_references_transitively vs proposed DRG walk — produce the resolve_transitive_refs contract in contracts/.
  • R-3: stringly-typed reference audit for dynamic dispatch of doctrine / curate / build_charter_context / etc. — produce the final "string occurrences that static grep missed" list.

D-9 — Phase 1 artifacts

Decision: Produced together with this plan — data-model.md, quickstart.md, and four contract files under contracts/.

Three sequential WPs. Each WP PR includes: source edits, test edits, its occurrence artifact, and a verifier-green CI run.

WP1.1 — Excise curation surface (tracks #476)

Scope:

  • Delete src/doctrine/curation/ (entire package)
  • Delete src/specify_cli/cli/commands/doctrine.py
  • Delete src/specify_cli/validators/doctrine_curation.py
  • Remove app.add_typer(doctrine_module.app, name="doctrine") from src/specify_cli/cli/commands/__init__.py
  • Delete src/doctrine/{directives,tactics,procedures,styleguides,toolguides,paradigms}/_proposed/ trees
  • Delete tests/doctrine/curation/** and tests/cross_cutting/test_doctrine_curation_unit.py
  • Update SOURCE templates under src/specify_cli/missions/*/command-templates/ and src/specify_cli/skills/ that reference any of the deleted surfaces (no agent-copy edits — agent dirs re-flow on upgrade)
  • Add scripts/verify_occurrences.py (reused by WP1.2/WP1.3)
  • Author kitty-specs/.../occurrences/WP1.1.yaml and seed kitty-specs/.../occurrences/index.yaml

Acceptance gates for WP1.1 PR:

  • All tests green
  • Verifier green against WP1.1.yaml
  • find src/doctrine -type d -name _proposed → no output
  • test -d src/doctrine/curation → non-zero
  • spec-kitty doctrine <anything> on a dev-install returns unknown-command
  • mypy --strict passes

WP1.2 — Strip inline reference fields from shipped artifacts, schemas, models (tracks #477)

Depends on: WP1.1 merged.

Scope:

  • Cross-check R-1 findings from research.md against src/doctrine/graph.yaml; add any missing edges to graph.yaml as an additive-only remediation before stripping inline refs
  • Strip tactic_refs: from 8 directive YAMLs, 3 paradigm YAMLs, 2 procedure YAMLs in src/doctrine/
  • Remove tactic_refs, paradigm_refs, applies_to field declarations from:
  • src/doctrine/schemas/directive.schema.yaml
  • src/doctrine/schemas/paradigm.schema.yaml
  • src/doctrine/schemas/procedure.schema.yaml
  • Remove corresponding fields from Pydantic models under src/doctrine/<kind>/models.py
  • Strip applies_to: list[str] from src/charter/schemas.py :: Directive
  • Update model tests in tests/doctrine/<kind>/test_models.py and consistency tests under tests/doctrine/ to assert the fields are absent
  • Author kitty-specs/.../occurrences/WP1.2.yaml and update index.yaml

Acceptance gates for WP1.2 PR:

  • All tests green
  • Verifier green against WP1.2.yaml
  • grep -R "tactic_refs\|paradigm_refs\|applies_to" src/doctrine src/charter/schemas.py returns zero hits outside explicit carve-outs listed in index.yaml
  • mypy --strict passes
  • WP1.1 acceptance gates still hold

WP1.3 — Validators reject inline refs + single context builder + reference-resolver excision (tracks #475)

Depends on: WP1.2 merged.

Scope (amended 2026-04-14 per internal review):

A. Baseline capture (must happen BEFORE any cutover step):

  • Before touching any source, capture the pre-WP03 rendered-text baseline for NFR-002b: run spec-kitty charter context --action <act> --json for each bootstrap action (specify, plan, implement, review) on current main and commit the JSON responses under kitty-specs/excise-doctrine-curation-and-inline-references-01KP54J6/baseline/pre-wp03-context.json (or one file per action). This file is the golden reference for the post-cutover byte-identity check.

B. DRG-backed helper + tests:

  • Add resolve_transitive_refs() to src/doctrine/drg/query.py per the corrected contracts/resolve-transitive-refs.contract.md — uses the live DRGGraph / Relation / walk_edges API; callers pass {Relation.REQUIRES, Relation.SUGGESTS} for legacy parity
  • Add src/charter/_drg_helpers.py + src/specify_cli/charter/_drg_helpers.py with _load_validated_graph(repo_root)
  • Add tests/doctrine/drg/test_resolve_transitive_refs.py including the 7-dimensional coverage in the contract and the behavioral-equivalence fixture against the still-present legacy resolver

C. Validator rejection:

  • Add InlineReferenceRejectedError (in src/doctrine/shared/errors.py or the existing shared-errors module)
  • Update per-kind validation.py (7 files: directives/tactics/procedures/paradigms/styleguides/toolguides/agent_profiles) to pre-scan raw YAML for tactic_refs/paradigm_refs/applies_to and raise the structured error
  • Procedures path (FR-008 amendment): the procedures pre-scan checks both top-level keys AND each steps[*] entry, since ProcedureStep.tactic_refs is a declared Pydantic field on current main (line 54). The field is removed from ProcedureStep in WP02 as part of the model edits; WP03 adds the pre-Pydantic scan so procedures get the structured error rather than Pydantic's extra_forbidden generic error.
  • Add tests/doctrine/test_inline_ref_rejection.py with one negative fixture per kind (7 total); the procedures fixture includes a step-level tactic_refs: entry to exercise the step-scan path

D. Resolver/compiler flip:

  • Swap resolve_references_transitively → resolve_transitive_refs in src/charter/resolver.py and src/charter/compiler.py (and their specify_cli/charter/* twins)

E. Context-builder cutover (strict D-1 sequencing):

  • Flip the five live build_charter_context() call sites to build_context_v2() (temporary intermediate state)
  • Rename build_context_v2 → build_charter_context, delete legacy implementation, update re-exports in both __init__.py files, flip the five call sites back to the renamed import
  • Add tests/charter/test_merged_graph_on_live_path.py asserting assert_valid() runs on bootstrap context builds (FR-016)
  • NFR-002b byte-parity check: spec-kitty charter context --action <act> --json for each bootstrap action must now byte-match the baseline captured in step A. Diff must be empty.

F. Legacy excision:

  • Delete src/charter/reference_resolver.py
  • Delete include_proposed parameter from src/charter/catalog.py :: load_doctrine_catalog() and update all callers

G. Test-coverage collapse (per D-4 — only after replacement green):

  • Rewrite tests/charter/test_context.py as a single-builder suite; include the NFR-002a artifact-reachability parity contract inherited from the soon-to-be-deleted test_context_parity.py
  • Add tests/doctrine/drg/test_shipped_graph_valid.py that calls assert_valid(merge_layers(load_graph(graph.yaml), None)) — subsumes the cycle-detection coverage currently in tests/doctrine/test_cycle_detection.py and tests/doctrine/test_shipped_doctrine_cycle_free.py
  • Delete tests/charter/test_context_parity.py
  • Delete tests/charter/test_reference_resolver.py
  • Delete tests/doctrine/test_cycle_detection.py (coverage rehomed to test_shipped_graph_valid.py)
  • Delete tests/doctrine/test_shipped_doctrine_cycle_free.py (coverage rehomed)
  • Update tests/doctrine/test_artifact_kinds.py — replace the three from charter.reference_resolver import _REF_TYPE_MAP uses with the equivalent doctrine.artifact_kinds.ArtifactKind public enum (or delete the affected test cases if they were purely about the private mapping)
  • Update tests/charter/test_resolver.py:293 — change the patch("charter.resolver.resolve_references_transitively", ...) target to patch("charter.resolver.resolve_transitive_refs", ...) with the new return-value shape
  • Update tests/charter/test_compiler.py:107 — comment reference to resolve_references_transitively becomes resolve_transitive_refs
  • Update tests/agent/test_workflow_charter_context.py for the single-builder name

H. Occurrence artifacts:

  • Author kitty-specs/.../occurrences/WP1.3.yaml and finalize index.yaml with the complete mission-level "zero occurrences" assertion; add resolve_references_transitively, ResolvedReferenceGraph, _REF_TYPE_MAP, _Walker to the must-be-zero set alongside the existing strings

Acceptance gates for WP1.3 PR:

  • All tests green
  • Verifier green against WP1.3.yaml and mission-level index.yaml
  • grep -R "build_context_v2\|reference_resolver\|resolve_references_transitively\|ResolvedReferenceGraph\|_REF_TYPE_MAP\|_Walker\|include_proposed" src tests returns zero hits outside explicit carve-outs
  • grep -R "curation\|_proposed\|tactic_refs\|paradigm_refs\|applies_to" src tests returns zero hits outside explicit carve-outs
  • NFR-002a: post-WP03 tests/charter/test_context.py inherits the artifact-reachability parity contract from test_context_parity.py and passes for every (profile, action, depth) case previously exercised
  • NFR-002b: spec-kitty charter context --action <act> --json for each bootstrap action byte-matches the pre-WP03 baseline committed under kitty-specs/.../baseline/pre-wp03-context.json
  • Test suite runtime within the 5% regression budget vs the pre-WP1.1 baseline (NFR-001)
  • mypy --strict passes

Risk & Mitigation Plan (operational)

Cross-reference spec Risks section. Operational mitigations:

Spec riskPlan-phase mitigation
Missing graph.yaml edges for inline tactic_refsPhase 0 R-1 produces the exhaustive inventory. WP1.2 adds missing edges before stripping YAML fields, atomically in the same PR.
Five-call-site inventory incompletePhase 0 R-3 runs static grep + Python AST walk + importlib/string-based dispatch search. The WP1.3 occurrence artifact must list every hit.
compiler.py / resolver.py carry non-resolver behavior we'd accidentally losePhase 0 R-2 produces the behavioral surface diff. Plan mandates "rewrite-not-delete" for both modules in WP1.3.
Per-kind validators drift during rewriteFR-014 test suite demands one negative fixture per kind (7 kinds). CI blocks on missing coverage.
Test deletion cascade breaks unrelated importsVerifier + mypy --strict + pytest collection run per PR.
NFR-001 runtime budget breachedMeasured at WP1.3 PR; reshape fixtures if needed.
Out-of-order WP mergeBranch-level sequencing + the occurrence artifact's requires_merged: [...] field gates the verifier.

Complexity Tracking

No Charter Check violations. No complexity budget items. The only architectural addition is resolve_transitive_refs() in src/doctrine/drg/query.py, which is a contract-equivalent replacement for an existing (deleted) function.

Branch Contract (restated)

  • Current branch at plan start: main
  • Planning/base branch: main
  • Final merge target: main
  • branch_matches_target: true

All three WP PRs open against main and merge back to main.

Next Step

User runs /spec-kitty.tasks --mission excise-doctrine-curation-and-inline-references-01KP54J6 to materialize the three WPs listed above.