The Agent Skills Directory

[COMMAND_EXECUTION]: Automates version control and repository management using the official git and gh (GitHub CLI) binaries. The skill includes extensive instructions for cloning, branching, committing, and managing pull requests.
[SAFE]: Implements mandatory safety guardrails, including requiring explicit user confirmation before force-pushing to main branches or deleting branches, and recommending diff previews before every commit.
[DATA_EXFILTRATION]: Mentions local credential paths such as ~/.ssh/github_token within documentation for troubleshooting authentication. This is a standard configuration pattern for the GitHub CLI and is presented as a reference for users rather than a silent access mechanism.
[PROMPT_INJECTION]: Ingests potentially untrusted data from repository content, such as pull request descriptions and issue comments, which could contain indirect prompt injections.
Ingestion points: Commands like gh pr view, gh issue list, and git log retrieve text from external contributors.
Boundary markers: No specific delimiters are used to wrap external content.
Capability inventory: The skill has the ability to write to the repository, merge code, and modify issue states.
Sanitization: No explicit sanitization of external text is performed before processing.

git-github