The Agent Skills Directory

[SAFE]: The skill provides well-defined instructions for a standard engineering task (git commit management). The logic is transparent and focuses on adhering to a widely recognized industry specification (Conventional Commits).
[COMMAND_EXECUTION]: The skill utilizes standard git command-line operations, including git status, git add, git diff, and git commit. These commands are used as intended for local repository management and do not include any patterns of privilege escalation or unauthorized access.
[INDIRECT_PROMPT_INJECTION]: The skill ingest data from the local repository via git diff to inform the commit message content.
Ingestion points: The agent reads output from git diff and git diff --cached (SKILL.md).
Boundary markers: None explicitly defined for the diff content.
Capability inventory: Uses git add and git commit to modify the repository state.
Sanitization: The skill relies on the agent's internal logic and a validation checklist to ensure the generated message matches the spec, but does not implement programmatic sanitization of the diff content. The risk is considered low and typical for development-assistive agents.

kernel-git-commit