---
description: Optional HiveMind project.json setup — infer from repo, ask only gaps
argument-hint: (no args — guidance only)
model: claude-opus-4-6
---

# HiveMind workspace setup (optional)

**There is no mandatory multi-step CTO onboarding.** The older “fill every field in `/hm-init`” flow is obsolete.

## What to do instead

1. **Read the repository** (`read_file`, `list_directory`, `glob_files`) and `.hivemind/project.json` when it exists.

2. **Infer** project name, stack, and phase from manifests (`package.json`, `README.md`, CI, etc.). Prefer facts from the filesystem over guesses.

3. **Ask the user only for gaps** you cannot infer — one or two questions at a time. In **Hivemind CLI**, use native tools:

   - **`pick_one` / `pick_many`** — single or multi-choice
   - **`form`** — multi-field terminal panels (text, yes/no, choices)

   Do **not** rely on IDE-only “AskUserQuestion”; those are not wired in Hivemind CLI.

4. **Write `.hivemind/project.json`** (and append to governance memory/decisions logs) **only after** values are agreed — never overwrite silently.

5. **If `.hivemind/project.json` already has a proper `meta.name`** (not the placeholder `"my-project"`), report status briefly and stop unless the user asks to change something.

## Optional checks

- `/hm-status` for session snapshot
- `/hm-focus <agent>` when working with a persona
