The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to run commands such as git status, git diff, git log, and git blame. While the skill document outlines sanitization rules for file paths and commit references, the interpolation of user-controlled input into shell commands presents an inherent risk of command injection.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from the git repository. • Ingestion points: Data from git log (commit messages), git status (branch names), and git diff/blame (file contents) are pulled into the agent's context. • Boundary markers: No specific delimiters or instructions are provided to the agent to treat command output as data only. • Capability inventory: Use of the Bash tool to read repository state. • Sanitization: The skill implements path and character validation to prevent shell injection but lacks mechanisms to sanitize or escape natural language instructions embedded within repository metadata.

git-status