The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from git diff outputs as described in SKILL.md. Boundary markers for the diff content are absent. However, the capability to modify files is strictly gated behind a human-in-the-loop process using AskUserQuestion. No sanitization is performed on the diff data, but the agent's actions are restricted to those explicitly approved by the user.
[COMMAND_EXECUTION]: The skill uses git fetch, git diff, and git branch commands to perform its review. These commands are standard for repository analysis and do not involve executing external or untrusted code.

branch-review-before-pr