werkrevisor/docs/prs/2026-10-11-PR#90-review-report.md

PR#90: tools/review-report Drafts a Review by werkrevisor and Codex for a Foreign Repository

WARNING: This document describes only the change applied in this PR. It may already be outdated once the next PR is merged. Historic PR documentation is not maintained along with new PRs; treat it as a snapshot.

Problem

hs.hsadmin.ng has no account ai-reviewer yet, and werkrevisor does not run there. The user wants the full review anyway, with the rules, Jev, the model review and Codex, started per PR from the command line, and posts the result under the own name after a check. For hs.hsadmin.ng #322 this took two commands and a report written by hand.

Non-Goals

• No write to Gitea: the user posts the report.
• No run without a human check of the findings.

Specification

Feature: a draft report of werkrevisor and Codex for one PR

Scenario#90.01: the report holds both reviews and asks for a check

So that the user posts only findings that a human checked.

• Given a PR of a Gitea without werkrevisor, e.g. hs.hsadmin.ng #322
• When the user runs tools/review-report with its URL
• Then report.md holds the result of werkrevisor, the criticality and findings of Codex and the usage of both, under a note to check each finding before posting

Verified by

• test_reviewreport.ReviewReportTest.test_the_report_holds_both_reviews_and_asks_for_a_check
• test_reviewreport.ReviewReportTest.test_the_werkrevisor_part_has_no_markers

Scenario#90.02: a failed review is named in the report

So that a missing part never looks like a review without findings.

• Given a werkrevisor run or a Codex run that does not complete
• When the user runs tools/review-report
• Then the report names the part as not completed, with its cause

Verified by

• test_reviewreport.ReviewReportTest.test_a_failed_review_is_named_not_left_out

Solution

• werkrevisor/reviewreport.py and tools/review-report run tools/dry-run --llm and codex.run in parallel, take the result comment of werkrevisor without its markers, and write build/review-report/<server>/<owner>/<repo>-<n>/report.md.
• Tokens and the values of the Codex login are removed from the report.
• The report warns when werkrevisor and Codex reviewed different commits, e.g. after a push between the runs.
• README: a paragraph below "Codex runner".
• Werkbaum: #model.codex.report; #model.codex.on and its nodes werkrevisor and werkator are deployed.
• Version 0.35.0.

Open Questions

None.

Additional Changes

• Found by the first Codex review of the runner on this PR: codex.checkout kept one mirror per owner and repository and fetched from origin, so a token of one Gitea could reach another Gitea with a repository of the same name. Each server now has its own mirrors and worktrees, and every fetch names the URL of its server; a test fails without the fix. Mirrors from before are not used any more and can be deleted.

Attachments

A trial on hs.hsadmin.ng #322 took 67 s for both reviews in parallel. werkrevisor gave ORANGE there, YELLOW in a run before, since the model review differs from run to run.