The Agent Skills Directory

[SAFE]: The skill is designed as a specialized tool for code auditing and security review. It provides clear guidance on identifying common vulnerabilities and prioritizing findings based on severity and confidence. The instructions include a detailed checklist in references/checklist.md that targets security risks such as injection, broken access control, and supply chain failures.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted code snippets, files, and diffs as ingestion points (SKILL.md Workflow). It lacks explicit boundary markers to delimit untrusted content from its own instructions and does not define specific sanitization methods. The skill encourages the reproduction of bugs via scripts, which constitutes a capability that could be targeted. However, as these elements are essential for its primary purpose of security auditing, and the skill provides instructions to identify such attacks in target code, the inherent risk is mitigated by the specialized nature of the task.

code-review-pro