exa

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.90). This skill explicitly performs semantic searches of the open web via mcp__exa__web_search_exa and then extracts full page content with WebFetch (see SKILL.md Decision Flowchart "Exa finds relevant sources" -> "WebFetch extracts full content" and references/api-endpoints.md describing search results with url/text), which clearly ingests untrusted, user-generated third‑party content (webpages, social media, GitHub) that the agent is expected to read and synthesize, allowing that external content to influence subsequent actions.