The library
Rendered from docs/specs/2026-09-26_k-docker-image-audit.md; the markdown file is the record.
Scorecard
| Dimension | Grade | Notes |
|---|---|---|
| Plan adherence | Acceptable | Matches amended K scope; T8 remains open under the documented contingency. |
| Scope discipline | Excellent | K implementation commits respect ownership and prohibited-file boundaries; no training added. |
| Test coverage | Acceptable | Saved parity and smoke artifacts support results; full-run completion remains outstanding. |
| Review compliance | Acceptable | Previous findings addressed; complete revised smoke-script rerun remains deferred. |
| Freeze integrity | — | Skipped: no P1/P2/P3 freeze hashes recorded. |
| Regression check | Acceptable | Ruff passes. Independently verified 50 valid, template-ordered scores and 50/50 matching PCM hashes. Full TSV logging has a named blocker. |
| Documentation | Acceptable | Required documentation and disclosure present; minor comment drift remains. |
| Overall | Acceptable |
Commentary
Plan adherence / Test coverage / Regression check — downgrade to Acceptable. T8 is unfinished: the cache contained 369 result rows at inspection, with no full TSV. The spec records exit 137, the resumed run, and deferred host-side logging. Complete the result addendum and submission log action, or record the specified hard-stop measurements.
Test coverage / Review compliance — limited to Acceptable. The corrected negative offline check has saved refusal evidence using real input. The complete revised smoke script has not yet been rerun. Saved positive smoke outputs are identical; independently recalculated image parity gives matching ranks and maximum difference 0.000142672, within tolerance.
Test coverage / Regression check — verification limitation. Pytest could not start because the sandbox has no writable temporary directory; Docker access was denied. The reported 419-test pass was not independently reproduced. Repository-wide Ruff passes now.
Documentation — cosmetic downgrade to Acceptable. docker/smoke.sh still says repeat outputs “must be byte-identical,” while its assertion and amended spec allow differences below
1e-6. Align the comment with the implemented tolerance.