The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes dynamic context injection (!) to execute shell commands such as git status, git diff, and uv run prek during the skill load phase. These commands are used to prepare the AI model's context with real-time data from the repository, ensuring the generated commit messages are accurate and relevant to the staged changes.\n- [PROMPT_INJECTION]: The skill includes an indirect prompt injection surface by processing external data from git diffs and user-provided notes.\n
Ingestion points: Data is ingested through git diff --cached and the $ARGUMENTS variable.\n
Boundary markers: User notes are encapsulated within <user_notes> XML-style tags to distinguish them from instructions.\n
Capability inventory: The skill has access to execute git commit and uv run prek.\n
Sanitization: No explicit sanitization of the input content is implemented, although the skill enforces a strict Conventional Commits output format which limits the impact of potential injections.

commit-staged