The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill uses the !command`` syntax in SKILL.md to execute shell commands including git fetch and git branch when loaded to populate its context with repository data. While these are standard git operations, they occur automatically upon initialization.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted branch names from the repository environment without sanitization, creating a risk for command injection. Ingestion points: Branch names are retrieved in SKILL.md through dynamic context commands (git branch, git for-each-ref). Boundary markers: No delimiters or protective instructions are used to wrap the branch data. Capability inventory: The skill requires the agent to execute shell commands including git branch -d <branch> and git rev-parse <branch> using these branch names. Sanitization: The branch names are not escaped or validated, which could allow a malicious branch name containing shell metacharacters to execute arbitrary code.
[COMMAND_EXECUTION]: The skill directs the agent to execute shell commands that modify the local git repository state, specifically deleting local branches based on the analyzed output of previous commands.

branch-cleanup