hadar01/github-agent

Autonomous PR review and issue-to-PR engineering with Claude — review every PR or auto-fix labeled issues, no local setup.

View on GitHub

Trust Signals

Scorecard Score
not yet scored
Maintenance Recency
Stale
License
None
namedescriptionrequireddefault
commandWhich pipeline to run: review | issue | triageyes
targetGitHub issue / PR / repo URL to act on. If omitted, it is derived from the triggering event (the PR URL on pull_request, the issue URL on issues).no""
anthropic-api-keyAnthropic API key. Store it as an encrypted repo/org secret.yes
github-tokenToken used for GitHub API calls. Defaults to the workflow token.no${{ github.token }}
post(review) Post the review back to the PR. true/false.notrue
comment(issue) Comment a link-back on the source issue after the PR opens. true/false.nofalse
fork(issue) Push to your fork and open the PR from there. true/false.nofalse
fail-on-request-changes(review) If true, a REQUEST_CHANGES / NEEDS_DISCUSSION verdict fails the check and blocks merge. If false (default), the review is advisory: it posts findings but never turns the check red.nofalse
max-costHard USD ceiling for the run (passed as --max-cost). Empty = project default.no""
node-versionNode.js version to run the agent with.no20
extra-argsAdditional raw CLI flags appended verbatim (e.g. "--label=bug --max=3").no""
namedescription
verdictReview verdict: APPROVE / REQUEST_CHANGES / NEEDS_DISCUSSION / UNKNOWN.