---

name: onboard-user description: Interview to learn about the user's preferences, working style, and communication needs. Results saved to global user profile.

User Onboarding Interview

You are conducting an onboarding interview to understand this user. The goal is to learn their preferences, working style, and how they want you to behave.

Interview Principles

1. Conversational, not bureaucratic — This is a dialogue, not a form. Ask naturally, follow up on interesting answers.
2. One topic at a time — Don't overwhelm. Ask about one area, listen, then move on.
3. Adaptive — If an answer implies something, confirm it rather than asking redundant questions.
4. Summarize periodically — Every 5-6 questions, briefly summarize what you've learned.
5. Confirm before saving — At the end, show the full profile and ask if it looks right.

Question Areas

Work through these areas conversationally. Skip questions if already answered or clearly not relevant.

1. Background & Experience

• How long have you been coding, and what's your primary area? (web, mobile, ML, systems, etc.)
• What languages and frameworks are you most comfortable with?
• Do you consider yourself more of a generalist or specialist?

2. Communication Preferences

• Do you prefer detailed explanations or terse responses?
• Should I explain my reasoning when I make choices, or just give you the answer?
• When I have a recommendation, do you want to see alternatives too, or just the best option?
• How do you feel about me asking clarifying questions vs. making reasonable assumptions?

3. Working Style

• Do you prefer to plan thoroughly before coding, or dive in and iterate?
• How much autonomy should I have? Should I ask before most decisions, or just make them?
• Should I make commits myself, or always ask first?
• How do you feel about me refactoring nearby code while fixing bugs?
• When you give me a task, should I stick strictly to it or flag related issues I notice?

4. Quality & Risk

• Are you more "move fast and break things" or "measure twice, cut once"?
• How important are tests to you? (Always test everything / Test critical paths / Tests slow me down)
• How do you feel about technical debt? (Never / Sometimes acceptable / Ship now fix later)
• Should I proactively flag potential issues I notice, or only speak up when asked?

5. Pet Peeves & Strong Preferences

• What do AI assistants do that annoys you? (Be specific — this helps me avoid it)
• Any coding patterns or practices you strongly prefer? (functional vs OOP, early returns, etc.)
• Any patterns or practices you hate?
• Anything I should absolutely never do?

6. Tools & Workflow

• Any specific tools I should know about? (task trackers, PR workflows, etc.)
• Any terminal or editor preferences that affect how I should work?

Adaptive Logic

• Senior developer → Skip basic explanation preferences, assume high autonomy
• Prefers terse → Keep your questions and summaries short too
• Mentions specific pet peeves → Acknowledge and confirm you'll avoid them
• Seems in a hurry → Offer to do an abbreviated version covering just the essentials

Interview Flow

1. Opening: "I'd like to learn how you work so I can be more helpful. This usually takes 10-15 minutes. Ready to start?"

2. Questions: Work through the areas above conversationally. Don't ask every question — adapt based on answers.

3. Periodic summaries: Every 5-6 questions, say something like: "So far I'm hearing: experienced full-stack dev, prefers terse responses, high autonomy, careful about quality. Sound right?"

4. Closing: Show the complete profile you'll save. Ask: "Does this capture you well? Anything to add or change?"

5. Save: Write to ~/.claude/meridian/user-profile.yaml

Output Format

Save to ~/.claude/meridian/user-profile.yaml:

Copy# Meridian User Profile
# Generated by user onboarding interview
# Last updated: YYYY-MM-DD

background:
  experience_years: <number>
  experience_level: junior | mid | senior | staff
  primary_domain: <web_fullstack | backend | frontend | mobile | ml | systems | devops | other>
  languages: [<list of languages>]
  frameworks: [<list of frameworks>]
  generalist_or_specialist: generalist | specialist

communication:
  verbosity: terse | balanced | detailed
  explain_reasoning: always | on_request | never
  show_alternatives: always | sometimes | never
  ask_vs_assume: ask_often | balanced | assume_more

working_style:
  planning_preference: plan_first | iterate
  autonomy: low | medium | high  # low=ask often, high=just do it
  commits: ask | auto_with_review | auto
  opportunistic_refactoring: yes | no | ask
  scope_expansion: strict | flag_issues | proactive

quality:
  risk_tolerance: conservative | moderate | aggressive
  testing_importance: high | medium | low
  technical_debt_tolerance: none | some | acceptable
  proactive_issue_flagging: yes | no

pet_peeves:
  - "<specific things to avoid>"

strong_preferences:
  - "<specific patterns or practices to follow>"

never_do:
  - "<hard constraints>"

tools:
  notes: "<any tool-specific notes>"

After Saving

Tell the user:

• "Profile saved to ~/.claude/meridian/user-profile.yaml"
• "This applies to all your projects. You can update it anytime with /onboard-user."
• "Now let's continue with what you were working on."