helpcrunch

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

• Third-party content exposure detected (high risk: 0.80). The SKILL.md shows the agent uses Membrane actions to fetch and act on HelpCrunch data (e.g., "get-chat-messages", "get-chat", "HelpCrunch Article", "create-message"), which are user-generated/untrusted third-party messages/articles that the agent would read and could materially influence follow-up actions.