skills/openai/skills/chatgpt-apps/Gen Agent Trust Hub

chatgpt-apps

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • Project Scaffolding Script: The skill includes scripts/scaffold_node_ext_apps.mjs, which automates the creation of a project structure. This script utilizes standard Node.js file system APIs to generate configuration files, a server implementation, and a widget template. It specifies well-known and versioned dependencies from the @modelcontextprotocol organization.
  • Trusted External Resources: The instructions consistently guide the user towards official OpenAI developer documentation and GitHub repositories for implementation examples. These references are integral to the intended development workflow and originate from trusted sources.
  • Indirect Prompt Injection Surface: The skill facilitates the creation of interactive widgets that ingest data from tool results. It acknowledges this attack surface by providing guidance on partitioning data between structuredContent and _meta, and advising that host-delivered data be treated as untrusted input within the widget UI.
  • Command Execution for Development: The skill describes workflows involving local server execution and the use of HTTPS tunneling services (such as ngrok) to connect local development environments to ChatGPT. These are standard procedures for testing agentic integrations and are documented within a developer-centric context.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 12:15 AM
Security Audit — agent-trust-hub — chatgpt-apps