The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external websites which presents a surface for indirect prompt injection where a site could attempt to influence the agent's behavior.
Ingestion points: Website content accessed via browser_navigate, browser_snapshot, and browser_evaluate in SKILL.md.
Boundary markers: None present to isolate website text from the agent's instruction context.
Capability inventory: The skill is limited to browser automation and generating a design report; it does not have access to sensitive files or shell command execution.
Sanitization: No specific sanitization of the scraped text is performed before it is included in the output report.
[DYNAMIC_EXECUTION]: The skill uses browser_evaluate to execute JavaScript snippets on the target website to extract design data.
Evidence: Static JavaScript functions are used to compute styles (e.g., getComputedStyle) and detect frameworks (e.g., checking for window.__NEXT_DATA__). This is a standard and appropriate use of the Playwright MCP for web analysis.

design-inspector