The Agent Skills Directory

[SAFE]: The skill consists entirely of instructional markdown files defining a specific style for git commit messages. It does not contain any executable scripts, binary files, or external network requests.
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to analyze external data, such as git diffs and recent commit history, to generate its output. While this is a standard ingestion point for potentially untrusted data, the skill's lack of actionable capabilities (such as file writing or network access) prevents this from being a functional vulnerability.
Ingestion points: references/generation.md and references/checklist.md direct the agent to examine git diffs and the results of git log (specifically the last 10 commits).
Boundary markers: The instructions provide a strict output template, which helps separate agent instructions from generated content, though no specific input delimiters are defined.
Capability inventory: The skill is restricted to text generation and does not include tools or scripts for system interaction.
Sanitization: No sanitization of the input git data is performed before processing.

kf-g-git-commit-japanese-commit-message