Skip to content
MCP AI Agent Guidelines

code-review

Cross-Modelworkflow

Mission

Section titled “Mission”

Inspect → grade → recommend → close the loop. Every review produces actionable findings.

When to Use

Section titled “When to Use”

Use when reviewing existing code for quality, security vulnerabilities, correctness, maintainability, API surface hygiene, or compliance adherence.

Triggers: “review this code”, “code review”, “check for security issues”, “quality review”, “audit this”, “grade this output”, “inspect this PR”

Skills Invoked

Section titled “Skills Invoked”

3× free parallel scan → strong synthesis:

  • qual-review (Zero-Cost) — style and correctness
  • qual-code-analysis (Zero-Cost) — static metrics
  • qual-performance (Efficient) — performance audit
  • qual-security (strong, independent) — OWASP Top 10, injection, auth
  • qual-refactoring-priority (Efficient) — debt ranking

Chain-To

Section titled “Chain-To”
  • policy-govern — escalate governance issues
  • code-refactor — proceed with refactoring recommendations
  • test-verify — add tests for coverage gaps

Example

Section titled “Example”
{
  "request": "Review the new payment processor integration for security and correctness"
}

Output: Finding table ordered by severity (critical → low), each with file path + line, description, and suggested fix. Security findings are always separated into their own section.