The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes standard git and git-crypt commands for repository encryption management, key distribution, and troubleshooting. These commands are typical for the tool's documented purpose.- [DYNAMIC_EXECUTION]: Fixed Python one-liners are used to safely verify the existence of the GITCRYPT header in staged or committed blobs. The script only reads a specific number of bytes to check for a magic string and does not evaluate arbitrary code.- [DATA_EXPOSURE_AND_EXFILTRATION]: Instructions explicitly forbid printing repository keys, private GPG keys, or decrypted file contents to the terminal. It provides methods to verify encryption state without revealing sensitive data.- [INDIRECT_PROMPT_INJECTION]: The skill operates on repository metadata and file contents. Potential risks are mitigated by instructions that prioritize non-interactive verification and explicitly warn against credential exposure during history cleanup or access management.
Ingestion points: File paths, .gitattributes rules, and Git object metadata.
Boundary markers: The instructions provide specific command templates for the agent to follow.
Capability inventory: Includes git CLI access and python3 for verification.
Sanitization: Emphasizes verifying encryption status without printing contents.

git-crypt