The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a wide range of standard git CLI commands to manage the repository lifecycle. This includes staging changes (git add), creating commits (git commit), inspecting history (git log, git show), and performing safety checks (git check-ignore, git diff --check). All commands are standard tools for a developer-oriented skill and are used with explicit safety warnings, such as avoiding git push unless requested.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the local repository, such as diffs and commit messages, which could theoretically contain malicious instructions. However, the skill implements strong mitigation strategies by enforcing multiple manual review gates before any action is finalized.
Ingestion points: Data enters the context via git status, git diff, git log, and by opening untracked files.
Boundary markers: The skill relies on structured CLI output formats and explicitly instructs the agent to read and reconcile the diff hunks manually.
Capability inventory: The skill possesses local write capabilities via git add, git commit, git switch, and git cherry-pick. It does not perform network operations by default.
Sanitization: The skill mandates the use of git diff --cached --check to identify whitespace and conflict markers, and requires a full manual re-verification of the staged diff before execution to ensure no out-of-scope content is processed.

vibe-commit