imagine-agentlisted

# Imagine Agent Also fires on 'imagine-agent' or any request where a non-technical client wants to imagine an AI agent for their product or repo and their plain-language goals need to become a buildable spec. ## What this skill does The **client-facing front door** to building a managed agent. It interviews a *non-technical* client in their own language — outcomes, users, what "good" looks like, never model names, tools, rubrics, or cron syntax — then **researches the target repo** to learn what's actually possible and what house standards exist, and **translates** the client's goals into the technical fields a builder needs. The output is a `build-sheet.json` it hands to `/build-agent`. Two skills, one seam: `imagine-agent` writes the build-sheet; `build-agent` reads it. See the contract in the sibling skill directory: `../build-agent/references/build-sheet.md`. ## The core move `/build-agent`'s own interview asks semi-technical things ("which sources? what schedule? which connector?"). **This skill answers those from research instead of asking the client.** A client goal like *"my users should get a summary every morning"* becomes — without the client ever seeing it — a model pick, a binary rubric, a cron deployment, the repo's already-wired Slack connector, and a dedup memory store. The only things that bounce back to the client are genuinely non-technical ("where do you want to receive it?"). ## Phase 1 — Interview the client (non-technical ONLY) Open warm and sho