Security-Focused Review
Review workflow specifically targeting security vulnerabilities
CLAUDE.md
When performing a security review:
- Check all user inputs: are they validated and sanitized at the boundary?
- Check database queries: are they parameterized? Any string interpolation in SQL?
- Check output rendering: is dynamic content escaped? Any use of innerHTML or dangerouslySetInnerHTML?
- Check authentication: does every protected endpoint verify auth? Are tokens validated correctly?
- Check authorization: does the code verify the user has permission for the specific resource, not just that they’re logged in?
- Check secrets: are there hardcoded keys, tokens, or passwords? Are secrets logged anywhere?
- Check dependencies: are there known vulnerabilities in new or updated packages?
- Check error responses: do they leak stack traces, internal paths, or system details?
Copy this workflow into your CLAUDE.md or agent config file so your agent follows this process automatically.