RonanCodes/setup-wizard

Setup Wizard

Interactive environment onboarding. Runs all sections by default, or target a specific one with flags.

Usage

/setup-wizard           # Full onboarding (all sections)
/setup-wizard --tokens  # API token setup only
/setup-wizard --ide     # IDE integration only
/setup-wizard --mcp     # MCP server setup only
/setup-wizard --check   # Verify everything works

Sections

Run each section in order (or just the flagged one). Use AskUserQuestion for every choice.

1. Plugin Installation

Detect what's already installed:

• Check ~/.claude/plugins/ for the ro plugin
• Run /plugin to see currently installed plugins

Recommended — install as a Claude Code plugin (bundles all skills, one command):

/plugin marketplace add RonanCodes/ronan-skills
/plugin install ro@ronan-skills

Skills then invoke as /ro:<skill-name> (e.g., /ro:ralph, /ro:commit).

Alternative for non-Claude-Code agents (Cursor IDE, other agents using SKILL.md):

• Cursor IDE: submit the repo via cursor.com/marketplace/publish or clone + point Cursor at skills/
• npx fallback (one-off, individual skills): npx skills add RonanCodes/ronan-skills/skills/<name> -g

1b. Repo ecosystem (self-register + recommend siblings)

This plugin is one of a family that hook together via ~/.claude/.env (resolved by scripts/resolve-repo.sh). On setup:

1. Self-register this plugin's location so sibling repos/skills can resolve it portably:

   bash "${CLAUDE_PLUGIN_ROOT:-$HOME/Dev/ronan-skills}/scripts/register-repo.sh" RONAN_SKILLS
   

2. Recommend the siblings that aren't installed yet:

   bash "${CLAUDE_PLUGIN_ROOT:-$HOME/Dev/ronan-skills}/scripts/ecosystem-check.sh"
   

   For each ✗ repo, offer to install it via the manifest's command (e.g. git clone the llm-wiki engine). ronan-private-skills is flagged Ronan-only — surface it as a pointer for Ronan, never as a default for others.

2. IDE Integration

Detect the environment:

• Check $TERM_PROGRAM — "vscode" means VS Code/Cursor
• Check for $CURSOR_* env vars — confirms Cursor
• Check for .cursor/ directory — Cursor project
• Check for .vscode/ directory — VS Code project

Based on detection, guide setup:

• Claude Code: skills go in ~/.claude/skills/ (global) or .claude/skills/ (project)
• Cursor: rules go in .cursor/rules/ (project) or .cursorrules (legacy)
• Both: set up both locations, explain the difference

3. MCP Servers

Check current MCP config:

• Read ~/.claude/mcp.json if it exists
• List configured servers and their status

Suggest useful MCP servers (ask which to install):

| Server | Purpose | Install | |--------|---------|---------| | Context7 | Up-to-date library docs | npx @anthropic-ai/claude-code mcp add context7 -- npx -y @upstash/context7-mcp@latest | | Playwright | Browser automation & testing | npx @anthropic-ai/claude-code mcp add playwright -- npx -y @anthropic-ai/mcp-server-playwright | | Fetch | HTTP requests from Claude | npx @anthropic-ai/claude-code mcp add fetch -- npx -y @anthropic-ai/mcp-server-fetch |

For each selected server, run the install command and verify it appears in config.

4. API Tokens

For each token, walk the user through setup:

| Token | Service | Signup URL | Used By | |-------|---------|-----------|---------| | PERPLEXITY_API_KEY | Perplexity AI | https://perplexity.ai/settings/api | perplexity-research | | OPENAI_API_KEY | OpenAI | https://platform.openai.com/api-keys | various | | GITHUB_TOKEN | GitHub | gh auth login or https://github.com/settings/tokens | GitHub integrations |

For each token:

1. Ask if they want to set it up (skip if not needed)
2. Explain what it's used for
3. Provide the signup URL or CLI shortcut
4. Ask user to paste the token (or run gh auth login for GitHub)
5. Verify it works:
   • Perplexity: curl -s -o /dev/null -w "%{http_code}" -H "Authorization: Bearer $KEY" https://api.perplexity.ai/chat/completions -d '{"model":"sonar","messages":[{"role":"user","content":"test"}]}'
   • OpenAI: curl -s -o /dev/null -w "%{http_code}" -H "Authorization: Bearer $KEY" https://api.openai.com/v1/models
   • GitHub: gh auth status
6. Save to ${CLAUDE_PLUGIN_DATA}/.env (Claude Code plugin) or ~/.config/ro/.env (other agents). Use mkdir -p "$CLAUDE_PLUGIN_DATA" && echo "KEY=value" >> "$CLAUDE_PLUGIN_DATA/.env".

4b. Background services & automations (opt-in)

The local factory ships optional background automations. None install by default — present the menu and let the user pick (AskUserQuestion, multiSelect). For each chosen item, explain what it does + its cost/footprint, then run its installer. Each must be easy to remove and say so.

Offer at least:

• Nightly dream maintenance (/ro:dream cron) — a 3am launchd agent that does safe, reversible factory upkeep (refines issues, fixes stale docs, drafts/archives skills) and leaves a morning report. Install: bash skills/dream/scripts/dream-cron.sh --install. Remove: same with --uninstall. Make clear it edits committed repos (undoable via git) and never deletes or contacts anyone external. Default off — many won't want an agent touching their repos overnight.
• Telegram two-way listener — a launchd daemon that pipes your Telegram replies into claude -p. Install: bash skills/telegram/scripts/setup.sh then the listener --daemon. Needs TELEGRAM_BOT_TOKEN + TELEGRAM_CHAT_ID (section 4).
• Pushover / Telegram notifications — just the tokens (no daemon); enables end-of-run pings. Covered by section 4; confirm here the notify scripts work.

For anything that runs unattended or reaches outside the machine, state the footprint plainly and require an explicit yes. After install, show the one-line status/remove command for each.

4c. Global engineering rules (~/CLAUDE.md)

Ensure the user's global Claude config carries the non-negotiable engineering rules. These are machine-wide behaviour, so they live in ~/CLAUDE.md (the "Global Claude Configuration" file), not a per-repo CLAUDE.md.

Check whether ~/CLAUDE.md already has an ## Engineering Rules (non-negotiable) section. If it is missing (or missing any of the three rules below), ask the user (AskUserQuestion) to confirm, then append/repair the section. Editing the user's global config is a structural change, so always confirm first; never overwrite unrelated content. The canonical block to ensure is present:

## Engineering Rules (non-negotiable)

These apply to every session, not just when a skill is loaded.

1. **Bugs are test-first.** When fixing a bug, FIRST write a test that reproduces it and fails (red), then make the fix that turns it green. Never patch a bug without a test that would have caught it.

2. **No human time estimates. Always pick the better solution.** You are a coding LLM, not a person on a deadline, so do not reason in human effort or down-scope to save time. Choose the genuinely better engineering solution unless the user explicitly asks for a quick hack or a time-box.

3. **NEVER use em-dashes or en-dashes.** Not in code, comments, commits, PRs, docs, or chat. Use commas, colons, parentheses, or full stops. `—` (U+2014) and `–` (U+2013) are both banned; do not substitute a hyphen for the intended punctuation.

If ~/CLAUDE.md does not exist, offer to create it with at least this section. Report what was added vs already present.

5. Check Mode (--check)

Verify all configured services still work:

• Run each verification curl/command from section 4
• Check MCP servers are responding
• Check skills are loadable
• Global rules: grep -q "Engineering Rules (non-negotiable)" ~/CLAUDE.md and confirm all three rules (test-first bugs, no time estimates, no em-dashes) are present.
• Background agents: launchctl list | grep ronanconnolly (dream, telegram-listener) — report loaded / not loaded.
• Report pass/fail for each item

Principles

• Never store tokens in plain text in SKILL.md or committed files
• Always verify a token works before declaring success
• If something fails, explain what went wrong and offer to retry
• Keep output clean — summarize results at the end of each section