people-data-labs

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

• Third-party content exposure detected (high risk: 0.70). The SKILL.md explicitly instructs the agent to fetch data from People Data Labs via Membrane (e.g., "membrane request CONNECTION_ID /path/to/endpoint" and actions like "enrich-person" / "retrieve-person"), which returns external person/social-profile data from public/untrusted sources (e.g., LinkedIn/social profiles) that the agent is expected to read and could materially influence subsequent behavior.