[JacobStephens2]

Summary

bug-origins → confirm / correct / reject
confirmed → exact Run attribution + proposed code-review Eval task
proposed Eval task → explicit approval → eligible for Experiments

Adds text/JSON views, retained settlement evidence, derived confirmed-origin counts in show --json, and private historical review repositories with draft catch criteria. Corrections and rejections withdraw superseded pending proposals, including reclosed bugs.

Closes #1029

Evidence

• Before: the rejection and confirmation CLI tests failed because settlement commands were absent; the Run-count test found no metric.
  After: cargo test --test bug_origins passes (15 tests; one ignored fixture entry).
• Before: each reviewer's exact merged-base, no-issue, and reclosed reproducer failed on its intended assertion.
  After: all three retained scenarios in tests/bug_origins/regressions.py pass individually and through the Rust integration test. Their assertions are preserved; the fixture additionally supplies the historical merge-commit metadata the corrected query reads.
• dotted-repo first failed because a valid repository name was invalid as an Eval folder name. The final fix hashes the complete bug identity into a valid task key; the fourth retained scenario, the full Bug-origin suite and strict Clippy pass.
• scripts/affected-tests d27c133274bd78d80f80d5f36c2b952c85233b18 passed: 50 site browser checks and 2,866 Rust tests across 68 test binaries (62 ignored fixture/platform entries). It used the cargo test fallback. Timed selector: 1,459.78 seconds, including compilation and browser checks. The final naming fix separately passes all Bug-origin tests and strict Clippy.
• cargo check, focused Eval/help/log tests, formatting and strict Clippy pass. CI passed on reviewed commit b4e2e0a; CI is rerunning for final commit 5b30b43.

Seams under test: the whole binary's bug-origins, scores, show --json and evals commands against fake GitHub and real fixture Git repositories.

Review coverage

Reviewed with thirdshift-code-review against fetched issue-978, pinned at d27c133274bd78d80f80d5f36c2b952c85233b18. Standards: one maintainability finding addressed. Spec: three behavior findings reproduced and fixed, with regression tests retained.

Both axes read all 22 changed files through b4e2e0a, including generated manual hunks and follow-up changes. Generated-output exclusions: none. No whole changed file was left unread. The final naming fix has post-review hunks not independently re-reviewed in src/evals/bug_origin.rs, tests/bug_origins/regressions.py, and tests/bug_origins/settlement.rs; these are the final review coverage gaps. Both reports retain the initial findings, final files-read lists and this disclosure in the Session's .thirdshift-review-oyMumX/standards.md and spec.md.

Unaddressed findings

Standards

None.

Spec

None.

Security

None.

Security review outcome

No unaddressed introduced findings.

Merge Danger

Door: two-way

Adds private settlement fields and Eval proposal folders; approval remains a separate Day-shift action.

Blast Radius: Scoring

Historical PR reconstruction and detailed local Run JSON.

Built with codex · gpt-6.1-sol · xhigh