exa-web-search-free

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

• Third-party content exposure detected (high risk: 1.00). The skill's SKILL.md and references/examples.md explicitly instruct use of web_search_exa and get_code_context_exa (and optional crawling_exa) to fetch content from public sites like the open web, GitHub, and Stack Overflow, which the agent is expected to read and use for research/decisions, exposing it to untrusted third-party content that could carry indirect instructions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill configures and calls a remote MCP at https://mcp.exa.ai/mcp (also used with query params as https://mcp.exa.ai/mcp?tools=...) which is registered via mcporter config add and invoked at runtime to execute the tool endpoints and thus runs remote code/controls the agent's behavior.