exa-web-search

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly directs agents to retrieve and ingest public web content via Exa's /search and /contents flows (e.g., exa.search_and_contents, exa.get_contents, and the cURL /search and /contents examples in SKILL.md and references/exa-api-guide.md), and to read and use that content to produce answers and drive tool-calling, which exposes the agent to untrusted third-party content that could carry indirect prompt injections.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill instructs adding a remote MCP server URL (e.g., https://mcp.exa.ai/mcp?exaApiKey=YOUR_API_KEY) which is fetched at runtime by Codex/Claude MCP to provision tools that directly affect agent behavior and capabilities, so this external URL is a runtime dependency that can control the agent's instructions.