chatlayer

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 using the Membrane CLI to run actions and proxy requests to the Chatlayer API (e.g., "membrane action run" and "membrane request CONNECTION_ID /path/to/endpoint"), which returns Chatlayer data such as conversations and user utterances (user-generated/untrusted content) that the agent is expected to read and can drive follow-up actions like update/delete/insert—creating a real avenue for indirect prompt injection.