The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external source code, which creates a potential surface for indirect prompt injection where malicious instructions could be hidden in the code being analyzed. \n- Ingestion points: The source_code input defined in skill.yaml allows arbitrary text or code to be provided to the agent. \n- Boundary markers: There are no explicit delimiters or instructions provided in the skill definition to treat the source_code input as purely data or to ignore embedded instructions. \n- Capability inventory: The skill identifies code smells and generates code snippets. No high-risk capabilities such as file system writes, network operations, or subprocess execution were detected in the provided files. \n- Sanitization: No evidence of input sanitization or validation of the source_code input was found.

refactoring-suggester