The Agent Skills Directory

[SAFE]: No security issues detected. The skill provides instructions for using refactoring tools with appropriate safety precautions.
[INDIRECT_PROMPT_INJECTION]: The skill operates on codebase files which could potentially contain malicious instructions, creating a surface for indirect prompt injection.
Ingestion points: Processes source code using refactor_tool (SKILL.md).
Boundary markers: Includes instructions to preview all affected locations before applying changes (SKILL.md).
Capability inventory: Utilizes apply_refactor_tool to execute file modifications (SKILL.md).
Sanitization: The workflow relies on manual previewing and impact analysis tools (get_impact_radius_tool, detect_changes_tool) to mitigate risks.

Refactor Safely