---

name: setup description: "Guided onboarding wizard for Ouroboros setup"

/ouroboros:setup

Guided onboarding wizard that converts users into power users.

Standalone users (Codex, pip install): Use ouroboros setup --runtime codex in your terminal instead. This skill runs inside a Claude Code session. For other runtime backends, the CLI ouroboros setup command handles configuration. For full install and onboarding instructions, see Getting Started.

GitHub Copilot CLI users: Run ouroboros setup --runtime copilot (after pipx install 'ouroboros-ai[mcp]' or uv tool install 'ouroboros-ai[mcp]'). Setup will:

1. Live-discover available models from the GitHub Copilot models API (uses gh auth token) and let you pick a default. A bundled fallback list is used when offline.
2. Write orchestrator.runtime_backend = copilot and llm.backend = copilot plus your chosen default into ~/.ouroboros/config.yaml.
3. Register the MCP server in ~/.copilot/mcp-config.json so the next copilot session can call ooo ... skills.

Hyphen Anthropic IDs that the Ouroboros defaults use (for example claude-opus-4-6) are auto-mapped at runtime to the dotted form Copilot CLI expects (claude-opus-4.6), so existing config files keep working when you switch backends.

Usage

ooo setup
/ouroboros:setup
/ouroboros:setup --uninstall

Note: Claude setup does two things:

1. Runtime configuration — selects the Claude Agent SDK profile on MCP 1.x
2. CLAUDE.md integration (optional) — per-project, adds an Ouroboros command reference block

It deliberately leaves ~/.claude/mcp.json untouched because marketplace plugin wiring owns that file. [claude] and its explicit [claude-sdk] alias use MCP 1.x. The plugin launches [mcp] in a separate MCP 2 process with the dependency-free [claude-cli] worker.

---

Setup Wizard Flow

When the user invokes this skill, guide them through an enhanced 6-step wizard with progressive disclosure and celebration checkpoints.

Python Runtime (Required)

Before running any shell snippet below, define this resolver in the same shell. It accepts only Python 3.12 or newer, prefers python3 and then python, and uses uv as the final fallback. Call ouroboros_python directly and quote every argument passed to it; the function preserves arguments and heredoc/stdin input. Only the probe and child interpreter discard inherited CPython path-selection overrides; the caller shell keeps its environment unchanged.

<!-- ouroboros-python-resolver:start -->

ouroboros_python() {
  if command -v python3 >/dev/null 2>&1 &&
    (unset PYTHONHOME PYTHONPATH PYTHONPLATLIBDIR PYTHONEXECUTABLE __PYVENV_LAUNCHER__; command python3 -c 'import sys; raise SystemExit(sys.version_info < (3, 12))') >/dev/null 2>&1
  then
    (unset PYTHONHOME PYTHONPATH PYTHONPLATLIBDIR PYTHONEXECUTABLE __PYVENV_LAUNCHER__; command python3 "$@")
    return
  fi
  if command -v python >/dev/null 2>&1 &&
    (unset PYTHONHOME PYTHONPATH PYTHONPLATLIBDIR PYTHONEXECUTABLE __PYVENV_LAUNCHER__; command python -c 'import sys; raise SystemExit(sys.version_info < (3, 12))') >/dev/null 2>&1
  then
    (unset PYTHONHOME PYTHONPATH PYTHONPLATLIBDIR PYTHONEXECUTABLE __PYVENV_LAUNCHER__; command python "$@")
    return
  fi
  if command -v uv >/dev/null 2>&1; then
    (unset PYTHONHOME PYTHONPATH PYTHONPLATLIBDIR PYTHONEXECUTABLE __PYVENV_LAUNCHER__; command uv run --no-project --quiet --python '>=3.12' python "$@")
    return
  fi
  printf '%s\n' 'Ouroboros skills require Python >= 3.12 or uv on PATH.' >&2
  return 127
}

<!-- ouroboros-python-resolver:end -->

---

Step 0: Welcome & Motivation (The Hook)

Start with energy and clear value:

Welcome to Ouroboros Setup!

Let's unlock your full AI development potential.

What you'll get:
- Visual TUI dashboard for real-time progress tracking
- 3-stage evaluation pipeline for quality assurance
- Drift detection to keep projects on track
- Cost optimization (85% savings on average)

Setup takes ~2 minutes. Let's go!

---

Step 0.5: Community Support

Before we begin, check ~/.ouroboros/prefs.json for star_asked. If not true, use AskUserQuestion:

{
  "questions": [{
    "question": "Ouroboros is free and open-source. A GitHub star helps other developers discover it. Star the repo?",
    "header": "Community",
    "options": [
      {
        "label": "Star on GitHub",
        "description": "Takes 1 second — helps the project grow"
      },
      {
        "label": "Skip for now",
        "description": "Continue with setup"
      }
    ],
    "multiSelect": false
  }]
}

• Star on GitHub: Run gh api -X PUT /user/starred/Q00/ouroboros, then merge {"star_asked": true} into ~/.ouroboros/prefs.json
• Skip for now: Merge {"star_asked": true} into ~/.ouroboros/prefs.json
• Other: Merge {"star_asked": true} into ~/.ouroboros/prefs.json

Create ~/.ouroboros/ directory if it doesn't exist. Preserve any existing keys such as welcomeShown, welcomeCompleted, and welcomeVersion when updating star_asked:

ouroboros_python - <<'PY'
import json, os
path = os.path.expanduser('~/.ouroboros/prefs.json')
os.makedirs(os.path.dirname(path), exist_ok=True)
try:
    with open(path, encoding='utf-8') as f:
        prefs = json.load(f)
    if not isinstance(prefs, dict):
        prefs = {}
except Exception:
    prefs = {}
prefs['star_asked'] = True
with open(path, 'w', encoding='utf-8') as f:
    json.dump(prefs, f, indent=2)
    f.write('\n')
PY

If star_asked is already true, skip this step silently.

---

Step 1: Environment Detection

Check the user's environment with clear feedback:

ouroboros_python --version
which uvx 2>/dev/null && uvx --version 2>/dev/null
which claude 2>/dev/null

For diagnostics, list uv-managed Python installations when uv is available:

uv python list 2>/dev/null | grep "cpython-3.1[2-9]"

The resolver already rejects system Python below 3.12 and provisions a compatible uv-managed Python when needed. This does not make the isolated [claude-sdk] and MCP 2 profiles import-compatible.

Report results with personality:

Environment Detected:
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Skill Python 3.12+         [✓] Resolver-selected
uv Python 3.12+            [✓] Available
uvx package runner         [✓] Available
Runtime backend            [✓] Detected

→ Full Mode Available (via uvx + uv-managed Python >= 3.12)

Decision Matrix:

Environment	Mode	Action
Python >= 3.12 + Claude CLI	Ready	Configure [claude] SDK/MCP 1 and skills
uvx + Python >= 3.12	MCP-capable elsewhere	Use a supported CLI-backed runtime setup for isolated ouroboros-ai[mcp]
Python < 3.12 only	Install needed	Run uv python install 3.12 then proceed
No package runner or Ouroboros package	Install needed	Install uv first, then proceed

If deps are missing and the user doesn't want to fix manually, recommend uv. Prefer package-manager paths over the vendor pipe-to-shell when the user's environment supports them (pipx > pip > brew > vendor one-liner):

Or install uv (recommended — handles deps automatically). Any one of:
  pipx install uv
  pip install --user uv
  brew install uv          # macOS / Linuxbrew
  curl -LsSf https://astral.sh/uv/install.sh | sh   # vendor one-liner (last resort)
Then re-run: ooo setup

IMPORTANT: Never install [mcp,claude], [mcp,claude-sdk], or [all,mcp] together and never write a direct ouroboros or python -m ouroboros MCP fallback. MCP 2 launchers must use an isolated uvx --isolated --python '>=3.12' --from 'ouroboros-ai[mcp]' ... or pipx run --spec 'ouroboros-ai[mcp]' ... process. Only [mcp,claude-cli] is supported because the CLI worker is out of process. Do not write an Ouroboros entry to ~/.claude/mcp.json; the plugin owns that registration.

If prerequisites are missing, show:

Ouroboros requires uvx (recommended) or the ouroboros package installed.

Quick install (< 1 minute) — install uv via any of:
  pipx install uv
  pip install --user uv
  brew install uv          # macOS / Linuxbrew
  curl -LsSf https://astral.sh/uv/install.sh | sh   # vendor one-liner (last resort)
Then:
  uv python install 3.12

Then re-run: ooo setup

Celebration Checkpoint 1:

Great news! You're ready for the full Ouroboros experience.

---

Step 2: MCP Profile Boundary

Show progress:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  Verifying Runtime Boundary...
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

The default Claude SDK profile stays on MCP 1.x. The plugin-owned MCP server
runs MCP 2 separately and selects the `[claude-cli]` worker.
This setup enables:

  Visual TUI Dashboard    [Watch execution in real-time]
  3-Stage Evaluation     [Mechanical → Semantic → Consensus]
  Drift Detection        [Alert when projects go off-track]
  Session Replay         [Debug any execution from events]

Do not create, update, or remove ~/.claude/mcp.json. Existing entries may be user-managed or belong to another compatible runtime. Explain that advanced MCP workflows require a host-managed isolated [mcp] launcher. The Claude marketplace plugin or another supported host setup owns that registration.

Celebration Checkpoint 2:

Runtime boundary verified! You can now:
- Use Claude-native ooo interview, seed, evaluate, and unstuck workflows
- Use the Claude SDK on MCP 1.x with isolated MCP 2 tools
- Keep the Claude SDK and MCP 2 dependency graphs conflict-free

---

Step 3: CLAUDE.md Integration (Optional)

Ask with clear value proposition:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  CLAUDE.md Integration
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Add Ouroboros quick-reference to your CLAUDE.md?

This gives you instant command reminders without leaving
your project context.

What gets added (~40 lines):
- Philosophy and pipeline overview
- Command routing table with lazy-loaded agents
- Agent catalog summary

A backup will be created: CLAUDE.md.bak

[Integrate / Skip / Preview first]

If "Preview first", show:

<!-- ooo:START -->
<!-- ooo:VERSION:0.53.0 -->
# Ouroboros — Specification-First AI Development

> Before telling AI what to build, define what should be built.
> As Socrates asked 2,500 years ago — "What do you truly know?"
> Ouroboros turns that question into an evolutionary AI workflow engine.

Most AI coding fails at the input, not the output. Ouroboros fixes this by
**exposing hidden assumptions before any code is written**.

1. **Socratic Clarity** — Question until ambiguity ≤ 0.2
2. **Ontological Precision** — Solve the root problem, not symptoms
3. **Evolutionary Loops** — Each evaluation cycle feeds back into better specs

```
Interview → Seed → Execute → Evaluate
    ↑                           ↓
    └─── Evolutionary Loop ─────┘
```

## ooo Commands

Each command loads its agent/MCP on-demand. Details in each skill file.

| Command | Loads |
|---------|-------|
| `ooo` | — |
| `ooo interview` | `ouroboros:socratic-interviewer` |
| `ooo seed` | `ouroboros:seed-architect` |
| `ooo run` | MCP required |
| `ooo evolve` | MCP: `evolve_step` |
| `ooo evaluate` | `ouroboros:evaluator` |
| `ooo unstuck` | `ouroboros:{persona}` |
| `ooo status` | MCP: `session_status` |
| `ooo setup` | — |
| `ooo help` | — |

## Agents

Loaded on-demand — not preloaded.

**Core**: socratic-interviewer, ontologist, seed-architect, evaluator,
wonder, reflect, advocate, contrarian, judge
**Support**: hacker, simplifier, researcher, architect
<!-- ooo:END -->

If Integrate:

1. Backup existing CLAUDE.md to CLAUDE.md.bak
2. Append the block above
3. Confirm successful integration

Celebration Checkpoint 3:

CLAUDE.md updated! You now have instant Ouroboros reference
available in every project.

---

Step 4: Quick Verification

Run verification with visual feedback:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  Verifying Setup...
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Check skills are loadable:

ls skills/ | wc -l  # Should show 12+ skills

Check agents are available:

ls src/ouroboros/agents/*.md | wc -l  # Should show 20+ bundled agents

Confirm the saved Ouroboros config selects the default Claude Agent SDK runtime on MCP 1.x while ~/.claude/mcp.json was not mutated by this setup. The dependency-free Claude CLI worker remains a distinct, explicit [claude-cli] selection for the isolated MCP 2 process.

---

Step 5: Success Summary

Display with celebration:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  Ouroboros Setup Complete!
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Mode:                     Claude Agent SDK (MCP 1.x)
Skills Registered:        15 workflow skills
Agents Available:         9 specialized agents
MCP Server:               Host-owned (config not mutated)
CLAUDE.md:                ✓ Integrated

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  You're Ready to Go!
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Start your first project:
  ooo interview "your project idea"

Learn what's possible:
  ooo help

Try the interactive tutorial:
  ooo tutorial

Join the community:
  Star us on GitHub! github.com/Q00/ouroboros

---

Step 5.1: Model Choice (Claude Code)

Before continuing to repository setup, give Claude Code users the same optional control over models without making it a requirement. Ask in the user's language; for Korean, use:

{
  "questions": [{
    "question": "설정이 완료됐어요. 기본 모델 설정으로 바로 시작할 수 있고, 모델은 언제든 나중에 바꿀 수 있어요.",
    "header": "모델 설정",
    "options": [
      {
        "label": "바로 시작하기 (권장)",
        "description": "기본 모델 설정으로 바로 작업을 시작해요"
      },
      {
        "label": "직접 모델 설정하기",
        "description": "단계별로 모델을 바꾸거나 목록에 없는 모델 ID를 입력해 고정해요"
      }
    ],
    "multiSelect": false
  }]
}

• 바로 시작하기: Continue to Step 5.5.
• 직접 모델 설정하기: Read and follow ../config/SKILL.md. In the local Claude Code harness, it opens the same settings UI in the user's browser at a temporary localhost address. They can reopen it any time with ooo config; this choice never permanently locks a model.

---

Step 5.5: Brownfield Repository Scan

Scan a root directory for existing git repositories and linked worktrees, then register them in the Ouroboros DB. This enables interviews to use brownfield context for existing projects.

Show scanning indicator:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  Scanning for Existing Projects...
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Looking for git repositories and worktrees up to two directories below the scan root.
Only repositories and worktrees reached directly by this depth-bounded walk are registered.
Local repos and repos with any remote name are eligible.
This may take a moment...

Implementation — use MCP tools only, do NOT use CLI or Python scripts:

CRITICAL — deferred-schema guard (prevents "Invalid tool parameters"): setup can call ouroboros_brownfield before and after a user-selection turn. A deferred schema loaded before scan is NOT guaranteed to remain loaded for the later set_defaults call. Immediately before EVERY ouroboros_brownfield call in this section, re-run tool discovery query: "+ouroboros brownfield" (idempotent — a no-op when already loaded). If the load returns no matching tool (and the tool is not already callable — an empty load for an already-exposed tool is an expected no-op, not absence), use the non-MCP setup fallback instead of retrying the failing call.

1. Load the brownfield MCP tool: tool discovery query: "+ouroboros brownfield"
2. Call scan+register:

   Tool: ouroboros_brownfield
   Arguments: { "action": "scan" }
   

   This walks scan_root up to two directory levels deep for valid seed repos/worktrees and registers them in DB. Each repo or worktree reached directly by the walk is registered self-only. Git worktree families are not expanded, so main or sibling worktrees outside the depth-bounded walk are not pulled in. Existing defaults are preserved.

Scan boundaries:

• The filesystem walk starts at scan_root; when omitted, scan_root defaults to the current user's home directory.
• Repositories are discovered directly by walking directories inside scan_root, at most two levels deep.
• Dot-prefixed directories and known noisy directories such as node_modules are not walked as seed locations.
• Both normal repos with a .git directory and linked worktrees with a .git file are registered when the walk reaches them.
• Git worktree families are not expanded. A worktree is registered only when the depth-bounded walk finds it directly.
• Local repos, repos without remotes, and repos whose remotes are not named origin are all eligible.

The scan response text already contains a pre-formatted numbered list with [default] markers. Do NOT make any additional MCP calls to list or query repos.

Display the repos in a plain-text 2-column grid (NOT a markdown table). Use a code block so columns align. Example:

Scan complete. 8 repositories registered.

 1. repo-alpha                   5. repo-epsilon
 2. repo-bravo *                 6. repo-foxtrot
 3. repo-charlie                 7. repo-golf *
 4. repo-delta                   8. repo-hotel

Include * markers for defaults exactly as they appear in the scan response. Do not summarize or truncate the list. The user needs to see all repo numbers to pick defaults.

If no repos found, skip the default selection prompt and proceed to Step 6.

Default repo selection — end the turn with the list:

Do NOT use AskUserQuestion for this selection. Assistant text emitted between tool calls is not guaranteed to render, so a question dialog fired in the same turn can appear without the repo list the user needs to answer it. Option preview fields cannot hold the list either — the preview box has a fixed height and silently truncates long lists.

Instead, end the turn with the repo grid as the final message so its display is guaranteed, and collect the selection as a plain chat reply.

Immediately below the grid, append the selection prompt:

If defaults exist:

Current defaults: <current default names> (numbers <current default numbers>)

Reply with repo numbers to change defaults (e.g. "6, 18, 19"),
"keep" to keep the current defaults, or "none" to clear them.

If no defaults exist:

No defaults set.

Reply with repo numbers to set defaults (e.g. "6, 18, 19"),
or "none" to run interviews in greenfield mode.

Then end the turn — no tool calls after the grid.

On the next turn, parse the user's reply:

• Numbers (any separator) → those indices
• "keep" (defaults exist) → skip the MCP call, confirm defaults unchanged, proceed to Step 6
• "none" → empty indices (clear all)
• Anything else → ask again in plain text; do not guess

Then re-run tool discovery query: "+ouroboros brownfield" and use ONE MCP call to update all defaults at once:

Tool: ouroboros_brownfield
Arguments: { "action": "set_defaults", "indices": "<comma-separated IDs>" }

Example: if the user picks IDs 6, 18, 19 → { "action": "set_defaults", "indices": "6,18,19" }

This clears all existing defaults and sets the selected repos as default in one call.

If "none" → { "action": "set_defaults", "indices": "" } to clear all defaults.

Celebration Checkpoint 5.5:

Brownfield defaults updated!
Defaults: podo-app, podo-backend, grape

These repos will be used as context in interviews.

Or if "none" selected:

No default repos set. interviews will run in greenfield mode.
You can set defaults anytime by running ooo setup again.

---

Step 6: First Project Nudge

Encourage immediate action:

Your first Ouroboros project is waiting!

The best way to learn is by doing. Try:

  ooo interview "Build a CLI tool for [something you need]"

Or explore examples:
  ooo tutorial

You're going to love seeing vague ideas turn into
crystal-clear specifications. Let's build something amazing!

---

Progressive Disclosure Schedule

Reveal features gradually to avoid overwhelm:

Immediate (Plugin Mode)

• ooo interview - Socratic clarification
• ooo seed - Specification generation
• ooo unstuck - Lateral thinking

After a Supported MCP Host Setup

• ooo run - TUI execution
• ooo evaluate - 3-stage verification
• ooo status - Drift tracking

Power User (Discover organically)

• Evolutionary loop and ralph persistence
• Cost prediction and optimization
• Session replay and debugging
• Custom agents and skills

---

Uninstall

When invoked with --uninstall:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  Ouroboros Uninstall
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

This will remove:
- Ouroboros block from CLAUDE.md

This will NOT remove:
- Plugin files in .claude-plugin/
- Any generated seeds or projects
- Your project data

Uninstall Ouroboros configuration? [Yes / No]

If Yes:

1. Leave ~/.claude/mcp.json untouched
2. Remove <!-- ooo:START --> to <!-- ooo:END --> block from CLAUDE.md (if exists)
3. Confirm: "Ouroboros plugin configuration removed. To remove plugin files, run: claude plugin uninstall ouroboros"

---

Setup Troubleshooting

"No compatible Python found"

Plugin mode works without a global Python when uv is on PATH. The skill
resolver uses a compatible python3, then python, then uv-managed Python >= 3.12.

If neither a compatible interpreter nor uv is available, install one:
  macOS: brew install python@3.12
  Ubuntu: sudo apt install python3.12
  Windows: python.org/downloads

"uvx: command not found"

uvx is recommended but not required. Alternative:

For the default Claude SDK runtime:
  pip install 'ouroboros-ai[claude]'

`[claude-sdk]` is an explicit alias. Use `[claude-cli]` only for the isolated
MCP 2 server worker. Do not combine `[claude]`, `[claude-sdk]`, or `[all]` with
`[mcp]`, and do not add a direct Python fallback to mcp.json.

"~/.claude/mcp.json conflicts"

Ouroboros will merge with existing MCP servers.
If you see issues, share your ~/.claude/mcp.json content and
we'll help you resolve conflicts.

---

Conversion Metrics Track

Track these checkpoints for conversion optimization:

• Started setup (skill invoked)
• Environment detected successfully
• MCP/Claude profile boundary explained
• CLAUDE.md integration accepted
• Verification passed
• Brownfield repos scanned and registered
• Default brownfield repo selected
• First project started (ooo interview)
• First seed generated (ooo seed)
• First execution completed (ooo run)

A fully converted user = all checkpoints passed

RFC #1392 State Breadcrumb Footer

Your final response MUST end with exactly one breadcrumb footer line:

◆ <current state> → next: <recommended action>

Derive <current state> from live session state via ouroboros_session_status when that MCP projection is available; otherwise derive it from this skill's actual outcome. Never use a linear Step N of M footer because Ouroboros is an evolutionary loop. When the next action is genuinely a choice, list 2-3 honest options in the next: clause. The breadcrumb line must be the last line of the response.