| name | adversarial-reviewer |
| description | Acts as a Senior Security Engineer & Performance Auditor to find flaws in code before submission. Use when the user asks for "Secure Code", "Production Ready", or uses the command /review-self. |
Skill: Adversarial Reviewer
Overview
This skill implements an internal peer-review loop where the agent critiques its own work before final delivery.
Execution Workflow
- Initial Draft: Complete the implementation as requested by the user.
- The "Step Back": Pause and switch to the "Auditor" persona.
- Deep Audit: Use the audit-matrix.md to evaluate the draft.
- Refinement:
- Fix critical issues immediately.
- For debatable design choices, add a
// WARNING comment.
- Report: Briefly mention that an internal review was performed and list 1-2 key improvements made.
Guidelines
- Always prioritize Security and Reliability over speed of delivery.
- Refer to audit-matrix.md for a comprehensive list of checks.