kvdb

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

• Third-party content exposure detected (high risk: 0.80). The SKILL.md explicitly instructs the agent to use Membrane to proxy requests to the KVdb API (e.g., "membrane request CONNECTION_ID /path/to/endpoint" and "membrane action run ..."), which fetches data from the third-party KVdb service (https://kvdb.io/api) and therefore exposes the agent to untrusted/user-generated content that it will read and act on.