MCP OAuth Cloudflare
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines tools that process untrusted user input, which represents a standard attack surface for indirect prompt injection.
- Ingestion points: The example tool in templates/index.ts accepts a 'message' parameter.
- Boundary markers: No explicit delimiters are used to wrap the input in the example implementation.
- Capability inventory: The skill environment has network and persistent storage access.
- Sanitization: The skill provides HTML sanitization for the OAuth approval UI but doesn't implement specific prompt sanitization for the tool inputs.
Audit Metadata