The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted content from Git repositories, which introduces a surface for indirect prompt injection.
Ingestion points: The skill uses the Bash tool to execute git log, git status, and git diff, which bring commit messages, branch names, and file contents into the agent's context.
Boundary markers: There are no explicit instructions or delimiters used to separate user instructions from data read from the repository.
Capability inventory: The skill is granted Bash, Write, Edit, and Glob permissions, which could be misused if the agent obeys instructions embedded in commit messages or file diffs.
Sanitization: No sanitization or validation of the Git output is described before the data is processed by the LLM.
[COMMAND_EXECUTION]: The skill relies on the Bash tool to perform its core functionality, executing various Git and GitHub CLI (gh) commands. While expected for a Git skill, this provides a high level of capability to the agent.
[SAFE]: The skill provides documentation on defensive Git hooks and .gitignore templates aimed at preventing the submission of sensitive information like API keys and environment files.

git