About review-spec
review-spec is a code-review skill for evaluating a specification against the referenced codebase. It helps teams find feasibility, design, reliability, security, and testing concerns, with prioritized, actionable findings.
Key features
- 01Compare specification with code
- 02Check architecture and feasibility
- 03Assess reliability, security, and performance
- 04Identify edge cases and testing gaps
- 05Prioritize actionable findings
Use cases
- 01Review a proposed implementation plan
- 02Check whether a spec matches existing code
- 03Find security or reliability risks before development
Install
Copy the command and run it in a terminal.
$
npx skills add https:/ / github.com/ ferueda/ agent-skills --skill review-specAfter installing, type/review-specin Claude Code to confirm it loaded.
Details
From the project READMEReview the given spec document by analyzing both the spec AND the referenced codebase.
Process
- Read the spec - Understand goals, phases, and proposed changes
- Explore the codebase - Read files and code mentioned in the spec
- Analyze patterns - Identify existing patterns, conventions, and architecture
- Validate the plan - Ensure proposals align with codebase reality
- Identify issues - Find gaps, risks, and improvement opportunities
Review Dimensions
Evaluate across these areas (focus on what's relevant):
- Architecture: Component boundaries, data flow, API contracts, separation of concerns
- Feasibility: Implementation complexity, technology trade-offs, effort estimation
- Reliability: Error handling, retries, idempotency, graceful degradation
- Performance: Bottlenecks, caching, query patterns, scaling approach
- Security: Auth, data protection, input validation, audit logging
- Edge Cases: Null handling, limits, timeouts, race conditions, partial failures
- Testing: Testability, integration strategy, rollback considerations
Output Format
For each finding:
### [Finding Title]
**Category**: Architecture | Feasibility | Reliability | Performance | Security | Edge Case | Testing
**Severity**: Critical | High | Medium | Low
**Section**: [Spec section or phase]
**Issue**: [Clear problem description]
**Recommendation**: [Specific, actionable change]
**Rationale**: [Technical justification]
Guidelines
- Verify against code - Don't trust the spec blindly; check actual implementations
- Follow existing patterns - Recommendations should align with codebase conventions
- Be specific - Reference exact files, functions, and line numbers
- Prioritize - Order findings by severity and impact
- Challenge assumptions - Question decisions that lack justification
Comments
Markdown supported