The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes repository data such as code diffs and status reports, which are external untrusted inputs that could contain malicious instructions.
Ingestion points: The agent reads repository state through git status, git diff, and local policy configuration files located within the project directory.
Capability inventory: The skill is capable of executing shell commands to stage, commit, and push changes to remote repositories.
Boundary markers: The instructions throughout SKILL.md and the reference files strictly mandate that the agent must obtain explicit user authorization before performing any mutating action.
Sanitization: The skill explicitly instructs the agent to perform a "Secret and sensitive content check" in references/clean-tree-checklist.md, looking for API keys, tokens, and credentials in the diff before committing.

dizzy-commit