The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute standard git commands such as git status, git diff, git add, and git commit. These operations are limited to the git CLI and include explicit instructions to obtain user approval before modifying the repository state or creating commits.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from git diffs to generate commit summaries. While this constitutes an attack surface where malicious code changes could attempt to influence the agent's output, the skill mitigates this by requiring the user to review and approve all generated commit messages and staged paths.
Ingestion points: Git diff output processed in SKILL.md.
Boundary markers: None explicitly implemented for diff content.
Capability inventory: File staging and commit creation via git commands in SKILL.md.
Sanitization: Includes specific instructions to identify and block the staging of sensitive files like .env, private keys, and credentials.

commit-summary