The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes find and grep to search the local directory for documentation matching specific keywords. This operation is limited to file discovery and does not involve executing the contents of the files or external communication.
[PROMPT_INJECTION]: The skill processes untrusted data from the repository's markdown and CSV files. While this creates an indirect prompt injection surface, the risk is minimal as the agent uses the data for classification rather than for control flow or sensitive operations. * Ingestion points: Documentation files (.md, .csv) searched in Step 1. * Boundary markers: None present. * Capability inventory: Bash, Read, WebFetch, AskUserQuestion. * Sanitization: None present.

keel-audit