The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local bash scripts (hotfiles.sh, ownership.sh, secret-scan.sh) to perform analysis. The scripts use standard Unix utilities like git, awk, and sed. Best practices such as variable quoting and the use of Bash arrays for command arguments are followed, which mitigates the risk of command injection through repository paths or metadata.
[DATA_EXPOSURE]: The tool is designed to read Git history, which is its primary function. While it scans for sensitive information like API keys and private keys, it includes a redaction mechanism in secret-scan.sh that masks the discovered secrets in the output to prevent further exposure in the agent's context.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from Git history (e.g., commit messages, author names) that could be used for indirect prompt injection.
Ingestion points: Data enters the agent's context from the output of the analysis scripts located in the scripts/ directory.
Boundary markers: The scripts provide structured output in Table, CSV, and JSON formats, but do not explicitly use delimiters or instructions to warn the agent about embedded instructions in the audited data.
Capability inventory: The skill is restricted to executing the provided bash scripts and standard Git operations.
Sanitization: The scripts do not perform specific sanitization of Git metadata against prompt injection before returning results to the agent.

git-repo-audit