The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local git commands to retrieve commit hashes and code differences.
Evidence: SKILL.md and code-reviewer.md contain commands such as git rev-parse, git diff --stat, and git diff to identify the range of changes to be reviewed.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it processes untrusted data (code changes) through an LLM subagent.
Ingestion points: The subagent in code-reviewer.md receives the output of git diff {BASE_SHA}..{HEAD_SHA}, which contains raw code and comments from the repository.
Boundary markers: The prompt template lacks explicit delimiters or specific instructions to the subagent to disregard natural language commands that might be present within the code being reviewed.
Capability inventory: The skill utilizes the Task tool to dispatch subagents and executes shell commands via the terminal to gather git metadata.
Sanitization: There is no evidence of sanitization or escaping of the diff content before it is interpolated into the instructions for the reviewer subagent.

requesting-code-review