The Agent Skills Directory

[SAFE]: No malicious intent or suspicious instructions were detected. The skill's operations are consistent with its stated purpose of improving code structure and module separation.\n- [COMMAND_EXECUTION]: The skill uses tools including mcp__serena__list_dir and mcp__serena__get_symbols_overview in SKILL.md to map out the codebase. These tools are used solely for inspecting project metadata and symbols for refactoring purposes.\n- [PROMPT_INJECTION]: An indirect prompt injection surface is present because the skill processes untrusted source code. Ingestion occurs via the mcp__serena tools mentioned in SKILL.md. Boundary markers and sanitization logic are absent, which theoretically allows malicious comments in the code to influence refactoring decisions. However, given the primary purpose and local scope of the capabilities (file movement and import updates), the risk is minimal.

refactoring-packages