Frozen document. Dates and status refer to its original source. Private resources and unreviewed images are unavailable.
PR metrics and Audits Dashboard code review · 30 September 2026
Fixed point: origin/main at 0e8ebf4. First reviewed head: 157a572. Scope: the PR lifecycle, Greptile/sub-agent telemetry, and Audits Dashboard change. The code-review skill ran its two axes independently: Standards used gpt-6-luna at medium reasoning; Spec used gpt-6-sol at medium reasoning. Their token and run totals enter the transcript ledger after their turns are captured; no separate totals are inferred here.
Standards
No hard violation of a documented repository coding standard was found. The Standards reviewer made two correctness notes, not Fowler-smell violations:
| ID | Finding at first reviewed head | Mitigation |
|---|
| STD-01 | Greptile collection silently stopped at the first 100 PRs, reviews, or comments. | Paginate list endpoints and check runs; add tests for multiple pages. |
| STD-02 | When GitHub refresh failed, live transcript rows replaced previously saved Greptile rows despite a message saying saved data remained. | Combine saved Greptile checks with current transcript review turns, label their different capture times, and test the fallback. |
Spec
| ID | Finding at first reviewed head | Mitigation |
|---|
| SPEC-01 | Only each PR's latest head was scanned for Greptile checks, losing review history from earlier revisions. | Capture checks for every PR commit and retain head-specific counts. The saved snapshot rose from 9 current-head checks to 25 historical checks. |
| SPEC-02 | The refresh fallback hid saved GitHub reviews. | Same fix as STD-02; source labels distinguish saved GitHub checks from live transcript turns. |
| SPEC-03 | The broad “review comments” count omitted PR discussion and written review submissions. | Count non-author inline, issue-discussion, and nonempty submitted-review bodies separately; present the total and breakdown without storing text. Historical Atlas total became 34: 33 inline and one discussion/written-review comment. |
| SPEC-04 | The architecture assessment linked Assess deep-module opportunities before product wave as a mitigation before any design or refactor was approved. | Remove the mitigation link and label the owner-selected candidate as exploration, with no mitigation yet. |
| SPEC-05 | Growing SVG charts would scale to panel width and make labels unreadable. | Keep the chart's intrinsic width and allow horizontal scrolling as the timeline grows. |
No material unrequested scope was found. Fix commit 38a9393 mitigates STD-01/02 and SPEC-01 through SPEC-05. Both original reviewers rechecked that head: Standards found both notes resolved and no new serious issue; Spec found all five findings resolved, then noted that the audit record lacked mitigation links. This record and the Audits snapshot now link each finding to the fix commit.
Twenty-two focused Python tests, JavaScript syntax checks, and the final local browser smoke passed. The browser check rendered 9 PR rows, 34 review comments, 2 audit runs, 9 findings, the historical review rows, and a synthetic 30-day chart with horizontal scrolling; it closed Chrome afterward. Greptile will review the published PR head separately.