v0.46.4

Fixed

• Version sources: patch bump so commits that touch canonical version files
  satisfy check-version-sources / pre-commit together with CHANGELOG.md.

v0.46.2

Fixed

• Modules: init 0.1.29 — patch bump so dev → main PRs satisfy --enforce-version-bump
  against origin/main when main already had 0.1.28 (adding integrity.signature alone is not
  enough; the module version must increase when the manifest is in the diff).

Changed

• CI / modules: pr-orchestrator.yml verifies bundled modules without --require-signature
  (checksum + --enforce-version-bump + --payload-from-filesystem), so PR heads and non-main
  contexts are not blocked by missing signatures during implementation. sign-modules.yml auto-signs
  --changed-only manifests on push to dev or main for non-bot actors (requires
  SPECFACT_MODULE_PRIVATE_SIGN_KEY), runs strict --require-signature verification in the same job,
  then commits and pushes chore(modules): auto-sign bundled manifests [skip ci] when needed; pushes from
  github-actions[bot] skip signing and only run strict verify. Reproducibility resets to
  origin/<branch> after verify so it targets the post-auto-sign tip.

v0.46.0

Added

• GitHub hierarchy cache sync (#492) for backlog metadata used in agent and automation workflows.
• Agent governance loading (#493): leaner, deterministic bootstrap for canonical rule documentation.

Fixed

• Tests / CI: marketplace install mocks accept install_module(module_id, InstallModuleOptions(...));
  dynamic script loaders register modules in sys.modules before exec_module (doc frontmatter and
  verify-bundle-published gates; Python 3.11 compatibility job).
• Code review follow-ups: safer DevOps/source-tracking edge cases, stricter parsers and validators, and
  reduced duplication in analysis and sync helpers (see commits on dev).

Changed

• Governance / OpenSpec: archived completed changes and aligned internal wiki maintenance notes.

v0.45.1

Changed

• Dependency install profiles: the default wheel is slimmer—CrossHair, Hypothesis, Ruff, Radon,
  and unused pins (python-dotenv, cffi) are no longer in core dependencies. Use
  pip install specfact-cli[contracts] for CrossHair + Hypothesis, or pip install specfact-cli[dev]
  for contributors. packaging is pinned explicitly for module installer / PEP 440 use.
• Smart-test baseline fallback: incremental smart-test runs now establish a full-suite baseline when
  no last_full_run cache exists (avoids a no-op incremental pass and misleading zero coverage).
• Pre-commit single-invocation overwrite handling: staged Python files are passed to the code-review
  helper in one batch so .specfact/code-review.json is not overwritten by multiple xargs processes.

Fixed

• Missing bundle UX: when workflow bundles are not installed, the CLI now reports the
  marketplace module (e.g. nold-ai/specfact-codebase for the code group) instead of
  Command 'code' is not installed, which was easy to confuse with the VS Code code CLI.

• Generated GitHub workflow (resources/templates/github-action.yml.j2): GitHub Actions if
  conditions now use ${{ … }} so annotations, PR comment, and fail steps evaluate correctly
  on GitHub (avoids mixed always() && / raw expression parsing issues).

v0.44.0

Added

• Clean-code principle gates (clean-code-01-principle-gates):
  • .cursor/rules/clean-code-principles.mdc restructured as a canonical alias for the
    7-principle clean-code charter (naming, kiss, yagni, dry, solid) defined in
    nold-ai/specfact-cli-modules (skills/specfact-code-review/SKILL.md).
  • Phase A KISS metric thresholds documented: LOC > 80 warning / > 120 error per function;
    nesting-depth and parameter-count checks active. Phase B (> 40 / > 80) explicitly deferred.
  • AGENTS.md and CLAUDE.md extended with a Clean-Code Review Gate section listing
    the 5 expanded review categories and the Phase A thresholds that gate every PR.
  • .github/copilot-instructions.md created as a lightweight alias surface that references
    the canonical charter without duplicating it inline.
  • Unit tests: tests/unit/specfact_cli/test_clean_code_principle_gates.py covering all
    three spec scenarios (charter references, compliance gate, LOC/nesting check).

v0.43.3

Fixed

• First-contact docs contract hardening:
  • strengthened README / docs/index.md / CONTRIBUTING.md alignment tests
  • restored explicit clickable modules-docs landing link validation
  • hardened docs parity checks against filtered Jekyll site.* tokens and safer URL-host assertions
• Contract robustness for utility helpers under symbolic execution:
  • src/specfact_cli/utils/optional_deps.py now fails closed on invalid import targets
  • src/specfact_cli/utils/acceptance_criteria.py now rejects pathological control-character inputs
    without regex exceptions
  • src/specfact_cli/utils/enrichment_parser.py now uses safe regex helpers/guards so
    hatch run contract-test passes CrossHair exploration for enrichment parsing paths
• OpenSpec/docs review remediation:
  • wrapped overlong proposal bullets and corrected list spacing in active change artifacts
  • added cross-repo first-contact traceability guidance for the core and modules docs split

Changed

• Tests:
  • added utility regression tests for invalid package names, pathological acceptance criteria, and
    control-character enrichment blocks
  • converted docs entrypoint file presence checks from import-time assertions to a module-scoped
    skip fixture for clearer test behavior in partial environments

v0.43.2

Added

• Documentation ownership frontmatter rollout:
  • scripts/check_doc_frontmatter.py
  • docs/.doc-frontmatter-enforced (rollout scope)
  • hatch run doc-frontmatter-check
  • Pre-commit hook integration
  • Guide: docs/contributing/docs-sync.md
  • Sample pages include the extended schema; use --all-docs for a full-site check.

Fixed

• Tests: Register doc-frontmatter shared fixtures via root tests/conftest.py
pytest_plugins (Pytest 8+ forbids pytest_plugins in nested conftests); scope env/cache
  isolation to test_doc_frontmatter paths only.
• OpenSpec (openspec/config.yaml): SpecFact code review JSON dogfood tasks require a fresh
  .specfact/code-review.json, remediation of review findings before merge, and TDD evidence.
  Freshness excludes openspec/changes/<change-id>/TDD_EVIDENCE.md from the staleness
  comparison so evidence-only edits there do not by themselves force a new review run.
• scripts/pre_commit_code_review.py: invoke review from repo root (cwd), emit JSON as above,
  and enforce a 300s subprocess timeout with TimeoutExpired handling so the hook cannot hang
  indefinitely.

Changed

• CI: Docs Review GitHub Actions workflow runs hatch run doc-frontmatter-check
  and includes doc-frontmatter unit/integration tests alongside tests/unit/docs/, with path
  filters for frontmatter scripts and helpers.
• Pre-commit specfact-code-review-gate: switched from specfact code review run with
  --score-only to specfact code review run --json --out .specfact/code-review.json, writing
  governed ReviewReport JSON under gitignored .specfact/ for IDE and Copilot workflows. The
  hook now prints severity counts on stderr, a short summary with the report path (absolute path
  hint), copy-paste prompts for Copilot/Cursor, and relies on verbose: true so successful runs
  still surface that feedback next to the nested CLI output.
• Documentation: docs/modules/code-review.md updated for the JSON report path and the pre-commit hook behavior.
• Tests: tests/unit/scripts/test_pre_commit_code_review.py and test_code_review_module_docs.py updated for
  --json/--out, repo-root cwd, and timeout handling.
• Doc frontmatter validation now uses a Pydantic DocFrontmatter model, stricter tracks glob checks
  (fnmatch.translate + re.compile after bracket/brace balance), and a reference page at
  docs/contributing/frontmatter-schema.md.

v0.43.1

Changed

• Packaging: Workflow slash-command prompts (specfact.*.md) are no longer duplicated in the core wheel; canonical copies live in specfact-cli-modules bundle packages under each bundle’s resources/prompts/. Install bundles (or use a dev repo checkout with resources/prompts/) for specfact init ide prompt export.
• IDE template drift checks on startup resolve source templates via the same installed-module discovery path as specfact init ide, not a single core resources/prompts directory inside the package.

v0.42.6

Fixed

• specfact init ide multi-source export writes prompts to a flat layout under the IDE export root (for example .github/prompts/ or .cursor/commands/) so editors and agents can discover specfact*.prompt.md (or equivalent) without per-source subfolders.
• Prompt catalog: core omits template basenames already provided by an installed module, avoiding duplicate exports when both ship the same filename.
• Re-export removes legacy per-source segment directories and prunes stale flat specfact* exports when the selected sources change.
• Tests: import pytest for MonkeyPatch annotations in init IDE prompt selection tests (Ruff F821).

v0.42.5

Added

• specfact init ide builds a prompt-source catalog from core (bundled or repo resources/prompts) plus installed modules across builtin, project, user, and marketplace roots; defaults to exporting all sources; supports --prompts for non-interactive selection (all, core, comma-separated module ids) and an interactive multi-select when multiple sources exist.
• IDE prompt exports are written under per-source subfolders (for example .cursor/commands/core/, .cursor/commands/<owner>__<module>/) so filenames stay collision-safe.
• Startup IDE template drift checks resolve exports under the namespaced layout (flat or nested).

Fixed

• VS Code / Copilot: chat.promptFilesRecommendations lists only prompt sources actually exported by init ide; selective --prompts no longer leaves stale .github/prompts/... entries from unexported modules.
• Integration tests: restore module discovery / installer paths after the command-audit temp-home scenario so later unit tests do not observe leaked marketplace module roots.