positioning-map

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

• Third-party content exposure detected (high risk: 0.85). Yes—runtime calls to mcp__claude_ai_Exa__web_fetch_exa and mcp__claude_ai_Exa__web_search_exa fetch public competitor web pages (hero/pricing/case-study text) and feed the extracted prose into the agent’s LLM context for synthesis.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly calls mcp web-fetch functions at runtime (e.g., mcp__claude_ai_Exa__web_fetch_exa(urls=["https://brightdata.com/", "https://apify.com/", "https://firecrawl.dev/", "https://scrapingbee.com/", "https://anysite.io/"]) and templates like "https://.com") to fetch external pages whose verbatim content (hero copy, pricing, changelogs, etc.) is then injected into the agent's context and directly controls the generated positioning report.