hadron-spawnlisted

# Spawn a Hadron Agent Create a new agent, give it a task, and launch it — all through the `hadron` CLI. The CLI handles auth and talks to the server; you never write tmux or curl directly. ## Gather intent first Before spawning, settle on: - **name** — short, human-readable (e.g. "Auth Refactor"). The server slugifies it into an id. - **task** — the one-line brief the agent boots with. - **group** — which column (run `hadron ls` to see existing groups; default is `Workers`). - **launch** — `claude` (default, an AI agent), `codex`, or `shell` (a plain terminal). - **cwd** — optional subdirectory under the workspace to work in. ## Spawn it ```bash hadron spawn "Auth Refactor" \ --group Workers \ --task "Extract the session-token logic out of middleware into its own module; add tests." \ --launch claude \ --start ``` - `--start` auto-launches the agent and types the task as its first message. Omit it to create the agent idle (the user starts it from the dashboard). - `--cwd subdir` scopes the agent to a directory inside the workspace (must already exist). - `--related a,b` links sibling agents so they're aware of each other. - `--artifact path/to/brief.md` attaches a file the new agent should read first. ## After spawning 1. Confirm it appears: `hadron ls`. 2. Tell the user the agent name + id and that it's running. 3. If this agent and the new one should coordinate, the link is mutual context — mention it. ## Notes - `--launch` is an enum (`claude|codex|shell