The Agent Skills Directory

[SAFE]: The skill defines a set of operational rules and workflows for an AI agent to refactor code without changing its external behavior. It emphasizes the importance of names, reducing duplication, and flattening control flow.
[PROMPT_INJECTION]: No malicious injection patterns, such as instructions to ignore safety filters or override system rules, were found. The instructions are focused entirely on the legitimate task of code refactoring.
[DATA_EXFILTRATION]: No network operations, hardcoded credentials, or sensitive file path accesses were identified. The skill operates on the code provided in the context.
[REMOTE_CODE_EXECUTION]: There are no patterns suggesting the download or execution of external code. The reference to shared/references/model-preview.md is a local file inclusion typical for internal configuration and does not pose an external threat.
[INDIRECT_PROMPT_INJECTION]: The skill processes external code, which represents an attack surface. However, the instructions are highly specific, requiring the agent to diagnose code 'smells' by name and preserve existing behavior, which significantly mitigates the risk of the agent executing instructions hidden within the code being refactored.

clean-code