Managed Agents

Guide the user from a rough idea or a concrete spec to a working Anthropic Managed Agents configuration. This skill is a primer + setup coach, not a code generator. Hand off actual implementation to Claude Code / Codex once the design is locked.

Authoritative docs: https://platform.claude.com/docs/en/managed-agents/overview Beta header required on every API request: anthropic-beta: managed-agents-2026-04-01

Mode selection

Check how the skill was invoked:

• No arguments / vague opener → run Mode A: Menu.
• Arguments or a concrete idea provided up front → run Mode B: Direct walkthrough.

Mode A — Menu

Present exactly these four options and wait for the user to pick one: