building-ai-agent-on-cloudflare
Fail
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: CRITICAL
Full Analysis
- [EXTERNAL_DOWNLOADS] (SAFE): The skill references standard and trusted Cloudflare development tools including the Wrangler CLI and the Cloudflare Agents SDK.
- [COMMAND_EXECUTION] (SAFE): Instructions include standard development lifecycle commands such as 'npm create cloudflare' and 'wrangler deploy' which are necessary for the skill's primary purpose.
- [INDIRECT_PROMPT_INJECTION] (SAFE): The provided code demonstrates a standard architectural surface for handling untrusted data from WebSockets. While it lacks sanitization and boundary markers in the simplified example, this is expected for educational templates.
- Ingestion points: 'onMessage' method in the 'MyAgent' class receives data via WebSockets.
- Boundary markers: None present in the simplified code example.
- Capability inventory: Access to SQLite storage ('this.sql'), task scheduling ('this.schedule'), and AI inference ('this.env.AI.run').
- Sanitization: No explicit sanitization is shown in the template code.
Recommendations
- Contains 1 malicious URL(s) - DO NOT USE
Audit Metadata