Close CUST-WP-0064 after the 2026-08-24 unassisted fire ingested
clay-borg, close CUST-WP-0065 now that all 120 active repos project a
classification, and close ADHOC-2026-08-25. Mark CUST-WP-0067 T02/T10
done (reverse relays already gone; work-record recovery lives on 0068).
Park the later no-checkout SBOM regression as CUST-IN-0015. Teach the
classification gate to use this host's checkout path.
Root-caused the xfail left in place at end of the test-coverage review:
git history (commit e0f9330, CUST-WP-0055 T02/T06/T07, 2026-07-08) shows
the agentic-resources exclude_repo: true -> path_prefixes narrowing was
deliberate, done to bring the repo's real source under the terminology
scan while still excluding the generated session_memory/.store/ tree.
The test was never updated to match and had been silently failing since.
Replaced the stale whole-repo-exclusion assertion with one that checks
the actual current contract: session_memory/.store/ stays excluded,
README.md (and everything else) is now in scope. No xfail marker needed
-- config was correct, only the test was wrong.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Critical review of CUST-WP-0060 (T03-T06) found: shipped mechanism proven
only by manual/live-repo runs, no repeatable test suite; tests/ never
wired to CI at all (pre-existing gap, not introduced here).
- tests/test_validate_work_records.py: 30 tests against the live canon
registry/schemas — classify() incl. all grandfathered legacy id
schemes, per-kind schema gates (intake open/closed/promoted, decision
prepared/resolved, engagement prepared), multi-doc yaml handling,
malformed-yaml-only-errors-if-id-registered, main() end-to-end via
subprocess (exit codes, template placeholders, --strict escalation,
terminal-record historical grace incl. the boundary case that grace
must NOT mask real enum violations), and the jsonschema-unavailable
fallback path (the exact failure mode that broke the first Forgejo
CI run before the runner-substrate fix)
- tools/validate_work_records.py: docstring said 'authoritative detector
is fix-consistency C-25' — wrong, it landed as C-31 (C-25..C-30 were
already taken); comment now correct
- .forgejo/workflows/python-tests.yaml: wires tests/ to CI for the first
time in this repo (apt python3/pytest/jsonschema/yaml on the
node:20-bookworm substrate, same pattern as work-records.yaml)
- tests/test_scan_workstream_terminology.py: found one pre-existing,
unrelated failure while establishing the CI-representative baseline
(agentic-resources allowlist entry no longer sets exclude_repo — a
policy question, not a bug this task should resolve silently); marked
xfail(strict=True) with the finding recorded so CI has a clean signal
and a silent 'fix' doesn't go unnoticed either
Local verification with apt-sourced deps (jsonschema 4.10.3, matching the
CI runner's package source, not just pip): 34 passed, 1 known xfailed.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Extend scan_workstream_terminology.py with allowlist loading, --apply-allowlist,
and --check-prose-gate for regression detection. Commit the T08 exclusion
config and unit tests; mark T01/T03/T08 done and activate the fleet workplan.
