The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to audit external files and pull requests provided by the user. Maliciously crafted instructions within the files being reviewed could potentially attempt to influence the agent's behavior during the analysis.
Ingestion points: External code files, environment configurations, and pull request data processed via the /security-review command in SKILL.md.
Boundary markers: Not present; the instructions do not explicitly wrap external content in delimiters or provide "ignore embedded instructions" warnings for the agent.
Capability inventory: The skill's instructions focus on analytical tasks; it does not contain inherent network exfiltration or file-write capabilities.
Sanitization: No specific input sanitization or filtering logic is defined for the reviewed content.
[COMMAND_EXECUTION]: The skill includes documentation and example shell commands for utilizing security tools like git-secrets, trufflehog, and trivy. These are established, well-known security utilities used for identifying vulnerabilities and hardcoded secrets.

security-review