The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (git diffs and repository file lists) to identify potential security risks.
Ingestion points: The scripts/security-scan.sh script ingests data using git diff and git ls-files to perform heuristic checks for sensitive files and high-entropy strings.
Boundary markers: Not applicable, as the ingested data is used strictly for pattern matching via grep rather than for prompt interpolation or instruction following.
Capability inventory: The scanning script is restricted to local read operations and pattern matching. It does not perform network operations, file writes, or dynamic code execution based on the ingested content.
Sanitization: The data is processed as plain text for regex matching; it is never executed or evaluated in a way that would allow for command injection or instruction override.

git-security-checks