parallel-search

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly exposes the agent to open web content: SKILL.md and app.py document and implement tools web_search_preview (performs web searches via the search-mcp endpoint) and web_fetch (fetches/extracts content from arbitrary HTTP/HTTPS URLs), which the agent reads and returns for further processing—allowing untrusted third-party pages to influence agent behavior.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill hard-codes and uses the MCP endpoint https://search-mcp.parallel.ai/mcp at runtime (required for the web_search_preview/web_fetch tools) and those tool outputs fetch and return external web content into the agent context, which can directly control prompts.