The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute git CLI tools (git add, git commit, git status) to manage the local repository. It proactively mitigates risk by defining a list of prohibited commands, specifically banning network-reaching operations such as git push, git pull, and git fetch, as well as destructive operations like git reset --hard.
[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection because it processes untrusted content from the repository tree to generate commit messages.
Ingestion points: The agent reads potentially attacker-controlled content from the local working tree via git status and git diff output (Workflow Step 1).
Boundary markers: The instructions lack explicit delimiters or warnings to the agent to disregard instructions found within the code diffs it analyzes.
Capability inventory: The agent possesses the capability to stage files and generate/execute commit commands based on the analyzed data.
Sanitization: No validation or sanitization is applied to the diff output before it is used as context for the agent's message generation.

git-commit