The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted code changes from git diffs, representing an indirect prompt injection surface where malicious instructions could be embedded in analyzed data. 1. Ingestion points: File diffs from git diff, git show, and gh pr diff are processed in Workflow Step 1. 2. Boundary markers: No delimiters or ignore instructions are specified for the untrusted diff content. 3. Capability inventory: The skill can read/write files within the .agent-security/ directory and report findings to the user. 4. Sanitization: No sanitization or escaping of the ingested code content is performed before processing.
[SAFE]: The skill operates entirely within the local repository environment to perform security audits. It uses legitimate developer tools and does not attempt external network connections, credential harvesting, or system persistence. The identified prompt injection surface is inherent to code analysis tools and is mitigated by the skill's limited capabilities.

security-scan