gocanvas

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

• Third-party content exposure detected (high risk: 0.80). This skill explicitly instructs the agent to fetch GoCanvas data (e.g., "List Submissions", "Get Submission", "List Forms" in the "Popular actions" section) and to proxy arbitrary API calls via "membrane request CONNECTION_ID /path/to/endpoint", meaning it will ingest untrusted, user-generated third-party content from GoCanvas that the agent is expected to read and which could influence its subsequent actions.