The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-controlled files (JS, TS, CSS, etc.) to generate refactoring suggestions. While this creates a theoretical attack surface for instructions embedded in code comments to influence the agent, the risk is mitigated by explicit 'Behavior preservation' rules.
Ingestion points: Target files matching supported extensions (.js, .ts, .vue, .svelte, .astro, .css, etc.) as defined in SKILL.md.
Boundary markers: Output is formatted into structured markdown blocks (Preview mode) as specified in SKILL.md.
Capability inventory: Read, Glob, Edit, and Write tools as defined in the allowed-tools section of SKILL.md.
Sanitization: No specific input sanitization is performed on the code content; however, the skill relies on the agent's internal reasoning and provided shared rules to ensure code integrity.
[EXTERNAL_DOWNLOADS]: The skill optionally utilizes the context7 MCP to fetch official framework documentation (e.g., React, Vue, Sass) to provide accurate refactoring suggestions based on the latest standards. This is a legitimate data enrichment step from a well-known service provider.

front-refactor