The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied code via the $ARGUMENTS placeholder. As a code-processing assistant, it possesses a standard attack surface where malicious instructions hidden in the code being refactored could potentially influence the agent's behavior. However, this is expected behavior for a refactoring tool.
Ingestion points: SKILL.md (via $ARGUMENTS)
Boundary markers: Absent; the skill relies on the agent's internal safety protocols.
Capability inventory: The skill instructions focus on code analysis, refactoring, and test verification, which typically involve reading/writing files and executing local build/test commands.
Sanitization: Absent; the skill is designed to process arbitrary code structures.
[EXTERNAL_DOWNLOADS]: The resources/implementation-playbook.md file provides configuration templates for various third-party tools and GitHub Actions, including those from GitHub, CodeRabbit, and CodiumAI. These are well-known, established services in the developer ecosystem and are presented as best-practice examples for automated code quality reviews.
[SAFE]: The skill's logic and resources are focused on educational content, design patterns, and performance optimizations. There are no signs of obfuscation, data exfiltration, or persistence mechanisms. Examples provided in the playbook use generic placeholders (e.g., 'localhost', 'root', 'smtp.gmail.com') rather than sensitive hardcoded credentials.

code-refactoring-refactor-clean