OpenAI Apps MCP
Pass
Audited by Gen Agent Trust Hub on Nov 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the creation of MCP servers that process untrusted data from tool arguments and HTTP requests. \n
- Ingestion points: Files like
src/index.tsingest data through JSON-RPC payloads and tool arguments. \n - Boundary markers: The templates incorporate Zod for schema validation, providing a layer of protection against malformed input. \n
- Capability inventory: The skill allows the agent to use powerful tools like
BashandWriteto create and deploy cloud workers, which necessitates careful oversight when following user-provided prompts to generate code. \n - Sanitization: Input validation is demonstrated using Zod in the provided server implementation examples.
Audit Metadata