agami-deploylisted

# agami deploy — prepare a self-host bundle the user ships to their own host You are preparing a **deploy bundle** on the user's machine so they can stand up the multi-user agami server (the HTTP MCP server with OAuth + admin) that their team's Claude connects to. The bundle pulls the **published image** (`ghcr.io/agamiai/agami-core`) — there is **no repo to clone and nothing to build**. You gather a few inputs, write the bundle locally, and hand off the cloud steps you can't do. The local mirror of this (single-player, no network, no auth) is `agami-serve` — if the user only wants agami in their own Claude Desktop, point them there instead. ## HARD RULES (load-bearing — a deploy handles secrets) 1. **Never ask for the admin password (or any secret) in chat — not even temporarily.** The password is typed by the **user** directly into the `agami.env` file (Phase 2 hand-off), exactly like `agami-connect` does for DB credentials. You never see it. 2. **Never put a secret on a Bash command line.** `prepare_deploy.py` takes only non-secret values; `deploy_preflight` generates the signing secret *into the file*. Hosts render Bash calls in chat. 3. **Username/password is the only auth this skill sets up.** Do **not** collect Google/Microsoft client id/secret. Social login ships free but is a manual `agami.env` step — point the user at the in-repo deploy README if they ask, and move on. 4. **No signup, no license key, no LLM/embedding key.** None are required; don'