OpenAI Codex security directory

OpenAI Developers

Published by OpenAI

High risk

Helps Codex build AI applications, agent projects and ChatGPT Apps. Use OpenAI APIs, the Agents SDK and Apps SDK with connected platform access. Scaffold and troubleshoot projects, run local agent deployments and prepare app-submission materials.

Security assessment

Plutonium assessed this plugin for permissions, capabilities, and security-relevant behavior.

  • Hosted key picker and local app testing use browser UI: Can programmatically control a browser, including navigation, form input, clicks, capture, or page inspection.
  • Agents SDK skill deploys services through Deployment Manager: Can deploy code, applications, containers, workflows, or infrastructure to a persistent environment.
  • Plugin scaffolds projects and writes secrets/configuration: Can read, create, or change files on the local filesystem or in a selected project.
  • Hosted picker creates OpenAI project API keys: Can create or change persistent records, files, projects, permissions, settings, or credentials.
  • Project key is persisted in a workspace env file: Uses credentials, OAuth grants, API keys, or tokens that may remain valid after the current session.
  • Agent and Apps workflows generate and execute application code: Can execute code, commands, scripts, unrestricted queries, or dynamic gateway actions.
  • Generated apps can consume paid OpenAI API credits: Can initiate purchases, paid jobs, payments, refunds, infrastructure spend, or other financial activity.

Available capabilities

This add-on exposes 5 tools or capabilities.

  • agents-sdk
  • build-chatgpt-app
  • chatgpt-app-submission
  • openai-api-troubleshooting
  • openai-platform-api-key

The interactive security report will load automatically.