The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from Git repositories, including commit messages, author names, and file paths, which are considered untrusted external inputs.
Ingestion points: The scripts hotfiles.sh, ownership.sh, and secret-scan.sh read output from git log and git diff.
Boundary markers: There are no specific delimiters or instructions informing the agent to disregard potential instructions embedded within the repository metadata.
Capability inventory: The skill's capabilities are restricted to executing local Git commands and text processing (awk, sed, grep). It lacks network access, file-writing capabilities (beyond temporary files), or privilege escalation vectors.
Sanitization: While the scripts do not escape data for structured output formats like JSON or CSV, the secret-scan.sh utility includes a safety feature to redact detected secrets in its reports.

repo-audit