Quickstart: Verifying CI Scoping Gate Reliability
No data model or API contracts apply (CI-config + arch-guard change).
1. Corpus data triggers + runs the blocking suite (#3008, SC-001)
fast-tests-corpus job; a corpus regression makes the quality-gate decision failure.
(previously it did not run at all) and the corpus job appears + is required.
- Path-filter test: a diff confined to
packs/built-in/**selects thecorpusgroup and the - Manual: open a scratch PR touching only
packs/built-in/**and confirm the quality workflow triggers
2. No false-trigger on lifecycle churn (#3008, SC-002)
corpus group. (Also notes.md/trace/**.)
- Path-filter test: a diff confined to
kitty-specs/<m>/status.events.jsonldoes NOT select the
3. Docs dead-link gate is diff-scoped (#3147, SC-003)
broken links PASSES the blocking gate; introduce a broken link in a changed file → it FAILS.
- With an untouched file elsewhere carrying a broken link, a docs PR whose own changed files have no
- The whole-tree scan (scheduled/full-run) still reports the untouched broken link.
4. Arch invariants stay green (SC-004)
PWHEADLESS=1 python -m pytest tests/architectural/test_ci_quality_path_filters.py tests/architectural/test_ci_collection_completeness.py -p no:cacheprovider -q
Both green with the new corpus group claimed (no suite left unclaimed; no skip/noqa).