you-research

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

• Third-party content exposure detected (high risk: 0.85). Outsider-authored free text can be ingested via the agent-led workflow’s runtime use of You.com MCP content tools (you-search/you-contents) to fetch and extract web page text (and user-provided URLs) into the agent’s LLM context; the skill docs explicitly instruct reading those sources as evidence.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly instructs the agent to fetch and rely on You.com runtime docs and an MCP server (e.g., https://you.com/docs/api-reference/research/v1-research and https://api.you.com/mcp?tools=you-search,you-contents), which the agent must query at runtime and use to shape scripts/requests, so remote content directly controls the agent's generated instructions.