The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes standard git and gh (GitHub CLI) commands to audit and perform repository maintenance. It explicitly instructs the agent to avoid dangerous flags like --force or destructive commands like git stash clear unless the user provides specific approval after reviewing an audit plan.
[DATA_EXFILTRATION]: Network activity is restricted to standard Git operations such as git fetch and git push --delete, which are necessary for the skill's stated purpose of pruning remote-tracking refs and cleaning up remote branches. These operations are performed against the user's configured remotes and require separate confirmation.
[PROMPT_INJECTION]: The skill does not contain instructions that attempt to override safety filters or bypass the agent's core instructions. Instead, it defines strict "Core Rules" that prioritize repository integrity and user control.
[SAFE]: The skill follows security best practices for Git automation by separating local and remote risks, performing sequential execution to handle failures gracefully, and utilizing detached HEAD states to ensure deletion checks are evaluated in a clean context.

vp-git-cleanup