The Agent Skills Directory

[SAFE]: The skill is designed for automated code reviews. It uses the Read, Grep, Glob, and a restricted Bash tool (limited to git commands) to analyze source code within the local repository environment.
[PROMPT_INJECTION]: The instructions include a detailed checklist for the agent to follow. The use of the word "CRITICAL" in markdown headers is instructional context for the agent regarding the importance of specific coding standards (e.g., Pydantic v2 syntax) and does not represent an attempt to override system safety guidelines.
[DATA_EXFILTRATION]: The skill lacks network-enabled tools and does not reference any external URLs or domains. All operations are confined to the workspace, ensuring that sensitive data is not transmitted externally.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it processes external content (file changes) to perform its review.
Ingestion points: Code changes and file contents are read using Read, Grep, and Bash(git *) tools as instructed in SKILL.md.
Boundary markers: The instructions do not specify explicit delimiters or "ignore" instructions for the agent when processing file content.
Capability inventory: Capabilities are restricted to reading local files and executing git commands; no network or administrative capabilities are present.
Sanitization: There is no explicit sanitization step for the code content being reviewed, relying on the agent's logic to distinguish code from instructions.

code-review