What problem does it solve? NanoClaw groups default to the Claude agent backend, and switching to OpenAI's Codex requires copying provider files from a separate branch, wiring barrel imports, pinning a CLI manifest entry, rebuilding containers, and configuring credentials. This Skill automates that entire installation and authentication workflow so Codex can run as a full agent provider alongside or instead of Claude. ## Core Features & Use Cases - Provider Installation: Copies the Codex payload (host, container, and setup files) from the providers branch, appends barrel imports, and adds the pinned @openai/codex entry to container/cli-tools.json. - Vault-Only Authentication: Runs an interactive walk-through that stores a ChatGPT subscription token or OpenAI API key in the OneCLI vault, keeping secrets out of .env files and containers. - Per-Group Provider Switching: Enables operator-driven switching via ncl groups config update --provider codex, with validation tests and a matching REMOVE.md for clean uninstallation. - Use Case: An operator wants one NanoClaw group to use Codex's app-server (with MCP tools and server-side session history) while other groups stay on Claude; this Skill installs the provider, authenticates it, and switches only that group. ## Quick Start Ask the agent to install the Codex provider and authenticate it, then switch a group with ncl groups config update --id <group-id> --provider codex.
