OpenAI Apps MCP
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references and includes templates for installing several standard development packages and CLI tools from the official NPM registry, including
@modelcontextprotocol/sdk,hono,zod, andwrangler. These are standard requirements for the stated purpose of building MCP servers and deploying to Cloudflare Workers. - [INDIRECT_PROMPT_INJECTION]: The skill provides an architecture for processing user-supplied data through MCP tool handlers. While this presents a potential surface for indirect prompt injection, the skill specifically mitigates this risk by providing examples of strict schema validation using Zod and advising on the use of tool parameters over less secure elicitation methods.
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill includes security-critical configurations, such as limiting CORS origins to
https://chatgpt.com. It also provides instructions for generating secure environment variables (e.g.,COOKIE_ENCRYPTION_KEY) using standard cryptographic tools rather than providing hardcoded secrets.
Audit Metadata