The Agent Skills Directory

[SAFE]: The skill provides detailed instructions for git operations and repository maintenance with a strong emphasis on security and best practices.
[EXTERNAL_DOWNLOADS]: The skill references the installation of standard developer tools like gh (GitHub CLI), glab (GitLab CLI), fj (Forgejo CLI), git-filter-repo, and gitleaks using official package managers and repositories.
[DATA_EXFILTRATION]: Provides comprehensive guidance on preventing secret leakage, including unstaging sensitive files and using git-filter-repo for history scrubbing.
[COMMAND_EXECUTION]: Outlines a controlled workflow for git commands, requiring explicit user approval for destructive operations such as hard resets or force pushes.
[INDIRECT_PROMPT_INJECTION]: The skill processes repository data like logs and PR comments. It mitigates risk through self-checks that enforce human-style commits and prohibit AI attribution.
Ingestion points: Reads git log, git diff, and forge-based PR/MR descriptions across all core workflow steps.
Boundary markers: Employs an 'AI Self-Check' to prevent AI-generated artifacts from entering the repository history.
Capability inventory: Performs git operations, file system modifications, and network API calls via forge-specific CLIs.
Sanitization: Recommends the use of git-filter-repo for scrubbing malicious or sensitive content from the history.

git