skyvern

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

• Third-party content exposure detected (high risk: 0.75). This skill is a browser-automation MCP integration that, at runtime, can fetch and read arbitrary web page content (public web pages/URLs) into the agent’s context for vision/LLM reasoning, which is outsider-authored free text.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill configuration points at an MCP runtime endpoint (https://api.skyvern.com/mcp/) which is contacted at runtime to execute browser automation commands (remote execution/control), so this external URL directly controls execution and is a required dependency.