The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute various shell commands, including Git operations (status, diff, log, add, commit, push) and project-specific scripts found in files like package.json, Makefile, or justfile. This execution is the primary purpose of the skill and is constrained by instructions to run only "fast" checks by default.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from the repository's history and configuration files, creating a potential surface for indirect injection.
Ingestion points: Git logs, file differences, and project-defined scripts (e.g., in package.json).
Boundary markers: None explicitly defined for isolating external data during processing.
Capability inventory: Subprocess execution for git and project-defined build/test tools.
Sanitization: The skill mitigates risks by instructing the agent to carefully read command output rather than relying solely on exit codes and to explicitly avoid staging sensitive files.
[SAFE_PRACTICE]: The skill contains explicit safety guidelines to protect the user's environment, such as blacklisting .env and secret files from being staged in Git, and requiring user confirmation before pushing code to remote branches.

ship