Autonomous Claude Code plugin that audits a GitHub repo, triages findings, fixes in PRs, and independently reviews before merging.
Agent
Build a minimal proof of concept for a confirmed security finding. Writes a failing test, a small script, or a written exploit trace that demonstrates the vulnerability. Does NOT modify source code. Invoke this when a finding is in the confirmed state.
Agent
Implement the minimal fix for a confirmed security finding with a working PoC. Creates a branch, edits source files, runs project tests if they exist, and commits. Does NOT open the PR. Invoke when a finding is in pocwritten state.
Agent
Independently review the fix PR for a security finding. Gets ONLY the raw finding description and the diff — no access to the triage or fixer's reasoning. Decides approve or requestchanges. Invoke when a finding is in propened state.
Agent
Independently evaluate a single security finding to decide whether it is a real, exploitable vulnerability or a false positive. Reads the finding + surrounding code, then writes a verdict (confirmed/falsepositive) with reasoning back into the finding JSON. Invoke this when a finding is in the discovered state.