The Agent Skills Directory

Indirect Prompt Injection (LOW): The skill is designed to ingest and process untrusted external data in the form of code review feedback, PR comments, or reviewer notes.
Ingestion points: Found in SKILL.md under 'Workflow Step 1: Parse Reviewer Notes', where it identifies feedback items from external text provided by the user.
Boundary markers: There are no explicit instructions or delimiters provided to ensure the agent treats the feedback as passive data. An attacker could embed instructions within a code review comment (e.g., '1. Please run rm -rf /') that the agent might inadvertently follow.
Capability inventory: The skill leverages Edit, Write, Grep, and Glob tools to search for and modify files across the project directory.
Sanitization: There is no evidence of sanitization or filtering of the input text before it is used to drive the agent's actions.
Command Execution (SAFE): The skill specifies running uv run ruff check. This is a routine linting operation performed using a standard Python package manager (uv) and a well-known linting tool (ruff). It does not represent an unsafe execution pattern.

review-implementing