| name | review-methodology |
| description | Professional review methodology — code review, security audit, architectural review, and kanban swarm verification. References codify Google's code review standards, OWASP audit patterns, and architectural assessment frameworks. |
| version | 1.0.0 |
| author | Hermes Agent community |
| license | MIT |
| metadata | {"hermes":{"tags":["review","code-review","security-audit","architecture-review","verification","quality"],"related_skills":["github-code-review","codebase-security-audit","systematic-debugging","requesting-code-review"]}} |
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:
- Understand intent — What is the change trying to achieve? Read the description, issue, or brief first.
- Assess correctness — Does it achieve its stated goal? Start here before considering style or elegance.
- Assess quality — Is it well-constructed for its purpose? Complexity, maintainability, test coverage.
- Assess risk — What could go wrong? Edge cases, security, regressions, operational impact.
- 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.