The Agent Skills Directory

[SAFE]: The skill establishes strict formatting and workflow guidelines for Git operations without introducing dangerous commands or external dependencies.
[DATA_EXPOSURE]: Includes explicit security protocols requiring the agent to audit code diffs for API keys, debug logs, and sensitive files (like .env) before allowing a commit, which is a defensive best practice.
[INDIRECT_PROMPT_INJECTION]: Implements a specific safeguard in the 'Non-Negotiables' section instructing the agent to ignore any AI instructions found within the code diffs, effectively mitigating indirect prompt injection risks from processed data.

git-commit