awml
Sign in

chrismdp / fix-review-findings

A failed review is fixed

Public ·
Version 2 of 3
  1. #3 · chrismdp · · 0.8Latest
  2. #2 · chrismdp · · 0.6Showing
  3. #1 · chrismdp · · 0.6
Fork
Who may see each partWhere each part runsYAMLChecksNo mistakes
publicanyone who reads the repositorythe failed checkthe check's findings, the patchand the branch's commits[pushed since, or tried twice][otherwise]the findings and the patchthe fix[a check fails][all pass]the commit the review read[touches Claude's files, or thebranch moved][touches a workflow][none of these]TriggerBug review or Security reviewfails on a pull request from abranch of this repository,opened by someone withwrite access or by the appLoosely coupled: it reads only the failed check and the findings the review posted in it. Builds and landing don't call it or wait for itLinked to chrismdp/issue-to-pull-request#tokenConnectionThe repository, through itsown Actions tokenLinked to chrismdp/issue-to-pull-request#appConnectionThe repository, through itsown GitHub AppCode · bring-inBring in the failed check'sfindings, the pull request'spatch against main, and thecommit the review readDataThe findings, the patch andthe commit they were foundonCode · sortSort by rule: has anyonepushed since the review, andhas the factory already triedtwice on this pull request?A person's push since the review means the branch has moved on, so the fix would be for old code: it stops, saying soCode · doSay in the status commentthat the findings need aperson, and whyCode · doPost the fix's status commenton the pull requestOutside actionPost or edit the fix's statuscommentAllowDataThe spec, guide, code,AGENTS.md and REVIEW.md,from the pull request's branchAI · draftFix the findings, Importantones and Nits, test first,changing none of ClaudeCode's own filesThe session runs as the build's draft does: in a sandbox, keeping the login token from its commands, loading no settings or servers from the tree, with a token that only readsDataThe fix, as a patch, with itsmessageCode · checkRun the tests, vet, theexamples and wasm parity onthe fixCode · sortSort by rule: does the fix touchClaude Code's own files or aworkflow, and is the branchstill at the commit the reviewread?Code · doAdd the fix to the pullrequest's branch as one morecommit, fast-forward onlyCode · doPut the fix in the statuscomment as a patch for aperson to apply, since the appmay not change a workflowOutside actionPush one more commit to thepull request's branchAllowThe app's push starts CI and the reviews again on the new commit, so a fix that brings its own Important finding fails its check and comes back here, up to two tries. The fix is always a new commit, never a rewrite of the branch