The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted repository data to generate commit plans and messages, and it follows instructions from a potentially attacker-controlled file at docs/agents/commit-tags.md.
Ingestion points: Reads file changes from the working tree and a configuration file (docs/agents/commit-tags.md) from the local repository.
Boundary markers: No specific delimiters are used to wrap repository content during summarization, and the skill is instructed to treat repo-local configuration as authoritative over defaults.
Capability inventory: Executes local git commands (add, commit, status). It does not perform network operations like git push.
Sanitization: No explicit sanitization or validation of the repository content or the override configuration file is mentioned.
[COMMAND_EXECUTION]: The skill utilizes git commands to manage the working tree.
Evidence: Explicit instructions to use git add and git commit for staging and committing files.

commit