constructive-frontend

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

• Third-party content exposure detected (high risk: 0.80). The skill's DynamicFormCard and useMeta hooks explicitly fetch and parse the Constructive "app-public" GraphQL endpoint (CRM_ENDPOINT / _meta and table queries) — untrusted, user-provided table schema and records are read at runtime and directly drive form structure, query/mutation names, and subsequent create/update/delete actions, so third-party content can materially influence agent behavior.