anthropics/launch-your-agent — reverse-engineered prompt

Build me a Claude Code skill that helps a founder go from an idea to a live Claude Managed Agent in one guided flow. I want it to start by interviewing me in plain English about what I am trying to build, whether it is an internal helper, part of a product, or something users interact with directly. Then turn that into a sensible version 0 plan, launch the agent in my own Anthropic account using a local env file for the API key, run a basic evaluation against my definition of done, make one or two improvement passes, and if it is a recurring job, set it up to run on a schedule.

When it finishes, give me a clear wrap up with an overview page, the exact launch payloads and script, an eval scaffold, and a short next steps plan for version 1 and version 2. Keep the flow mostly automatic, but pause for anything important, and look up the current managed agent docs online if you need to.