crawl

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly crawls arbitrary public websites and returns page text/chunks (see SKILL.md "Quick Start" and "For Context: Agentic Research" examples) and the scripts (scripts/crawl.py and scripts/crawl.sh calling tavily_request and saving item["raw_content"]) instruct the agent to ingest that untrusted third‑party content into LLM context, enabling indirect prompt injection.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill makes runtime calls to the Tavily MCP endpoint (https://mcp.tavily.com/mcp) / Tavily crawl API to fetch arbitrary webpage content which is then returned to be injected into LLM context (directly controlling prompts), so the remote URL is a required runtime dependency that can influence agent instructions.