chrismdp / review-a-pull-requestA pull request is reviewedPublic · Version 1 of 2#2 · chrismdp · Yesterday 2:28pm · 0.8Latest#1 · chrismdp · 27 Oct 2026, 9:20pm · 0.6ShowingForkWho may see each partWhere each part runsYAMLChecksNo mistakespublicanyone who reads the repositorythe pull requestthe patch and main's filesits findingsits findingsTriggerA pull request from a branchof this repository, opened bysomeone with write access orby the app, is opened orpushed toA fork's pull request, or one opened by anyone without write access, starts nothing here: its author could steer the sessions, which hold the maintainer's login tokenLinked to chrismdp/issue-to-pull-request#tokenConnectionThe repository, through itsown Actions tokenCode · bring-inBring in the pull request'spatch against main, and main'sREVIEW.md and reviewpromptsA patch that touches Claude Code's own settings, .claude/, .mcp.json or a CLAUDE.md, is refused before any session starts, and no session loads settings or servers from the tree The reviews read the patch as data and follow main's prompts and REVIEW.md, so a branch can't rewrite what it's held to2DataThe patch, with the pullrequest's title and descriptionDataThe spec, guide, code,AGENTS.md and REVIEW.md,from mainAI · checkReview the patch for bugs,with fresh context, keepingonly the findings that survive asecond lookAI · checkReview the patch for security,in a separate session,alongside the bug reviewCode · doOnce both reviews are done,post each as a check on thehead commit: passed if itfound nothing, failed if it foundsomething Important, neutral ifonly Nits or Pre-existing faults,with its findings rankedFindings that carry anything like the login token are refused, never uploaded or posted An Important finding fails its check, so the change takes a fix, or the maintainer's deliberate bypass, to merge. Nits and Pre-existing faults only mark it neutral2Outside actionPost the Bug review andSecurity review checksAllowFlow only+−Fit