What this gives your agent
A structured review procedure that keeps findings evidence-based, ranks them by severity, and separates defects from optional style suggestions.
What it can do
- Identify behavioral regressions
- Rank findings by severity
- Check tests against risk
- Produce file and line references
What it needs
- Code diff
- Repository context
- Test results
What it returns
- Prioritized findings
- Open questions
- Residual test risk
Ask it to work
“Review this branch before merge.”
“Find regressions in the billing changes.”
