SuperAPI AI Directory

review-methodology

Testing & quality

View repository

Updated 10-16 07:07

About this skill

Review Methodology

Professional review standards for code, security, architecture, and swarm output verification.

Review Types

Type Reference When to load
Code review references/code-review-standards.md Reviewing a PR or diff — 9 dimensions from Google's engineering practices plus project-specific conventions
Security review references/security-review.md Auditing for vulnerabilities — threat modeling, vulnerability classes, dependency analysis, supply chain
Architectural review references/architectural-review.md Evaluating design decisions — coupling/cohesion, scalability, data flow, abstraction boundaries
Swarm gate references/swarm-verification.md Evaluating swarm worker outputs — passing or blocking the kanban verifier gate

Templates

Template When to use
templates/code-review-response.md Writing a PR review with findings organized by severity
templates/security-finding.md Documenting a security vulnerability with reproduction steps
templates/swarm-verdict.md Passing or blocking the kanban verifier gate with evidence

The Review Mindset

Every review follows the same arc, regardless of type:

  1. Understand intent — What is the change trying to achieve? Read the description, issue, or brief first.
  2. Assess correctness — Does it achieve its stated goal? Start here before considering style or elegance.
  3. Assess quality — Is it well-constructed for its purpose? Complexity, maintainability, test coverage.
  4. Assess risk — What could go wrong? Edge cases, security, regressions, operational impact.
  5. Communicate — Clear findings, severity levels, actionable next steps. Not personal, not dismissive.

Never skip to quality or style before understanding intent and correctness. A well-styled solution to the wrong problem is still wrong.