cloudflare

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

• Third-party content exposure detected (high risk: 0.90). The skill's references (e.g., references/agents-sdk/patterns.md and configuration.md) show an Email Processing w/AI workflow where the agent's onEmail handler reads arbitrary incoming email.text() (untrusted, user-generated content) and then uses AI summaries to trigger follow-up actions (e.g., scheduling sendAutoReply), so third‑party content is ingested and can directly influence agent behavior.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill contains runtime code that registers and queries an external MCP server (example URL https://mcp.example.com via env.MCP_SERVER_URL) — e.g., this.mcp.registerServer(...) and this.mcp.getAITools(...) — which fetches tools injected into the agent's model context and can directly control prompts/behavior.