Review Agent
What This Demonstrates
Section titled “What This Demonstrates”The Copilot Review Agent reviews pull requests for correctness, security vulnerabilities, and maintainability — prioritizing critical issues and suggesting concrete fixes.
Prerequisites
Section titled “Prerequisites”- A pull request open on the repository
- GitHub web UI or VS Code with the PR checked out
Prompts
Section titled “Prompts”Prompt A — Full Review
Section titled “Prompt A — Full Review”Review this PR for correctness, security, and maintainability. Prioritizecritical issues first, then medium/low findings. Suggest concrete patch-levelfixes.What to look for:
- Severity-based ordering (critical → medium → low)
- Concrete code suggestions (not just “consider fixing this”)
- Security findings (injection, auth bypass, secret exposure)
Prompt B — Targeted Review
Section titled “Prompt B — Targeted Review”Focus this review on API contract consistency (`success/data/error/meta` shape),error handling, and TypeScript strictness across touched files.What to look for:
- API response shape validation against the project standard
- Missing error handling (unhandled promise rejections, missing try/catch)
- TypeScript
anyusage or missing type annotations
Prompt C — Infrastructure Review
Section titled “Prompt C — Infrastructure Review”Review Terraform and Kubernetes changes for least-privilege, secret handling,and safe defaults. Flag anything risky for production.What to look for:
- Overly permissive IAM roles or NSG rules
- Hardcoded secrets or missing Key Vault references
- Missing resource limits or unsafe defaults
Follow-Up Prompts
Section titled “Follow-Up Prompts”- “Summarize the 3 most important changes I should make before merging.”
- “Generate a commit that addresses the critical findings.”
- “Compare this review with what CodeQL would catch.”