[antonwolfy]

Summary

Points from the reviews of #43, #44, #59, #69 and #75 that were neither changed nor answered before those PRs merged. Only tests and text change here, no package code.

Changes

Tests

• feat(questions): write questions.v2 (question), keep reading v1 (text) #43: questions refine reads its input without validating it. A set whose duplicate record has "topics": "perf", which strict reading refuses, is repaired and written. Reading it strictly instead passed the suite.
• refactor(novelty): the bundle record owns its validity #44:
  • shorten turns a bundle block it cannot slice (a file row without end) into an exit that names the atoms file, not a traceback.
  • An entry-point record without a stored outside_share reports the share its counts give (83%, where main before refactor(novelty): the bundle record owns its validity #44 printed 0%). That figure was printed but pinned nowhere. The gate figures for the other stored shapes are already pinned in test_novelty_bundle_guards.py. The parity golden file is a capture of main before refactor(novelty): the bundle record owns its validity #44, so these cases stay out of it.
• feat(arms): keep the judge's treatment material in the artifact #69:
  • The cross-judge --judges gate refuses one damaged question among eight when the ids are ints. A gate that read only str ids passed before.
  • from_artifact skips a non-dict arm next to one it serves.
  • unreproducible_arms skips records, arms, evaluations and scores of the wrong type.

Docs

• docs(readme): how to review a skill or a repository with dokimi #59:
  • The README says to add --all-atoms to the tier command and to run --estimate-only first, since --all-atoms alone drops the tier.
  • It names the credential variables repo reach passes on: ANTHROPIC_API_KEY, or AWS_BEARER_TOKEN_BEDROCK with AWS_REGION and CLAUDE_CODE_USE_BEDROCK.
• feat(arms): keep the judge's treatment material in the artifact #69, ci: parallel tests, coverage on 3.14 only, pip cache, one static job #75: re-flow the 128-character line in evaluating-treatments.md, the runner comment in ci.yml (111 characters) and the coverage bullet in CONTRIBUTING.md.

Risk

None to behaviour. No changelog fragment: nothing under dokimi/dokimi/ changes.

Testing

On a Linux review host:

• make ci-local: 19 passed, 0 failed, 4 skipped (the Docker catalog smoke, test-latest, docs deploy, semgrep). 5,118 tests passed, 1 skipped.
• Each new test kills its mutant:
  • load_questions(args.questions) without validate=False in refine;
  • except KeyError instead of ValueError around entry_point_only in shorten;
  • isinstance(question_id(r), str) in the gate;
  • a dropped isinstance(arm, dict) in from_artifact;
  • a dropped isinstance(s, dict) in unreproducible_arms.

Checklist

• Tests added/updated and pytest passes
• ruff, mypy, and schema checks pass
• Docs updated (or N/A) — see the docs-drift checklist in CONTRIBUTING.md
• No ad-hoc run artifacts committed
• New files carry the SPDX license header (or are exempt in .github/.licenserc.yaml)

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 90.14%. Comparing base (9753cc7) to head (3e276e4).
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #87      +/-   ##
==========================================
+ Coverage   90.06%   90.14%   +0.07%     
==========================================
  Files         205      205              
  Lines       24193    24193              
  Branches     3892     3892              
==========================================
+ Hits        21790    21808      +18     
+ Misses       1724     1713      -11     
+ Partials      679      672       -7     

see 4 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.