The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes content from git diff and git status, which are derived from the local filesystem. Malicious content within the files being committed could theoretically influence the generated commit message.
Ingestion points: SKILL.md instructions to run git status and git diff during the analysis phase.
Boundary markers: None explicitly defined in the subagent prompt template.
Capability inventory: Performs git commit via local command execution after user confirmation.
Sanitization: Includes a mandatory human approval step ("Present for Approval") where the user reviews and can edit the generated message before the commit is executed.
[COMMAND_EXECUTION]: The skill executes standard git commands (status, diff, commit) on the local system. These operations are limited to the scope of git repository management and are triggered by user interaction.

commit