The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from external websites and landing pages, which introduces a risk of indirect prompt injection where malicious instructions embedded in the audited site could attempt to influence the agent's behavior.
Ingestion points: The skill analyzes "rendered pages" and screenshots of external websites.
Boundary markers: The skill includes a specific section, "References are data, not instructions," which explicitly tells the agent to disregard any instructions found within external content.
Capability inventory: The skill itself primarily generates reports and does not execute dangerous system commands or write files directly, relying on the user to approve subsequent fixes.
Sanitization: No technical sanitization is mentioned, but natural language instructions provide a security boundary.
[EXTERNAL_DOWNLOADS]: The skill references an optional external MCP (Model Context Protocol) server at https://uxkin.com/mcp to provide additional design system data.
Evidence: The skill instructs the user how to manually add this server if desired for "UI materials" searches. This domain is owned by the skill author.

web-review