The Agent Skills Directory

[SAFE]: The skill defines a standard development workflow for documentation and version management. It outlines how to categorize changes (patch, minor, major) and provides templates for changeset files.
[INDIRECT_PROMPT_INJECTION]: The skill reads repository data using git status and git diff. While processing external data is a potential attack surface for indirect prompt injection, the skill includes a mandatory human-in-the-loop confirmation step ('Show the changeset text... and get their confirmation before committing') which significantly mitigates the risk of the agent performing unintended actions based on the processed data.

gen-changesets