zgagnon

Use when reviewing a test harness for adequacy — post-implementation (adversarial review), as an audit-loop reviewer node, or pre-spec to verify that planned coverage reaches every acceptance criterion. Classifies tests as outside-in vs unit, and flags structural blind spots that let green tests miss real regressions.

Use when reviewing code for security concerns — token handling, injection surfaces, auth paths, secret exposure, and permissions scope. Invoked during adversarial review post-implementation, as an audit-loop reviewer node, or on PRs that touch auth/secrets/deploy surfaces. Distinct from the built-in `/security-review` slash command (which runs an automated scan); this skill guides a human or reviewer-agent through the audit.

Use when reviewing an implementation against its spec, plan, or written brief — during adversarial review post-implementation, as an audit-loop reviewer node in a pipeline, or at spec-time to self-audit for internal consistency. Maps each spec clause to code evidence and flags drift.