convex

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

• Third-party content exposure detected (high risk: 0.80). The skill includes an HTTP router that reads and processes webhook payloads (http.route path "/webhook" calling request.json()) and a public file-upload flow where clients POST arbitrary files to generated upload URLs, meaning the agent ingests and is expected to process untrusted third-party/user-provided content.