agent-skills-audit
por swader
Web & Frontend DevelopmentResumen
Run a two-pass, multidisciplinary code audit led by a tie-breaker lead, combining security, performance, UX, DX.
Descripción
Qué hace esta skill
Runs a structured two-pass code audit using six specialist roles: security, performance, UX, DX, edge case, and a tie-breaker lead. Each role analyzes the target codebase independently, then a reconciliation pass resolves conflicts and merges findings into one report ordered by severity, blast radius, and exploitability. File references, reproduction steps, and regression test suggestions are included for each finding.
Replaces ad-hoc single-angle reviews with one structured pass covering security, performance, UX, DX, and edge cases simultaneously, producing a single prioritized report instead of fragmented per-discipline feedback.
Cuándo usarla
- Auditing a pull request before merging to production
- Generating a risk-ranked remediation plan for a legacy service
- Finding security and performance gaps across API endpoints
- Stress-testing a codebase for edge cases and invariant violations
- Pre-launch review of backend and frontend code
Flujo de trabajo de ejemplo
User asks: Audit my Express API codebase for security vulnerabilities and performance issues
- 1
Read code and product flows to identify assets, entry points, trust boundaries, and high-risk operations
- 2
Map critical invariants across all mutating paths including HTTP routes, webhooks, and async jobs
- 3
Run specialist reviews in sequence: security, performance, UX, DX, edge case master
- 4
Tie-breaker lead resolves conflicts, removes duplicates, and assigns severity and confidence to each finding
- 5
Run cross-review pass 2 with specialists reassessing prior and edge-triggered findings
- 6
Publish final report ordered by severity with file references, concrete fixes, and verification tests
Prioritized findings report with severity ratings, file references, reproduction steps, a remediation plan, and an executive summary
Instalación
clawhub install swader/agent-skills-auditEjecuta el comando de la CLI o pega el prompt en Claude Code, Cursor, Codex o el agente de programación que uses.