/user-background-interviewer

name: user-background-interviewer
description: Conducts a deep background interview to understand what types of app ideas this specific user is most likely to succeed with based on their domain knowledge, networks, and skills.

<!-- version: 0.2.0 | outputs: memory/user_profile.md -->

Skill: user-background-interviewer

Purpose

Understand what types of ideas this specific user is most likely to succeed with. Surface hidden advantages, domain expertise, and natural distribution channels the user may not have considered.

Input

• Fresh conversation (no prior context needed)
• Optional: existing `memory/user_profile.md` to extend

Process

Step 0 — Check for Existing Profile

Before showing the opening message, check if `memory/user_profile.md` already exists.

**If a profile exists**, present a summary and ask:

> I already have a profile on file for you: > > - **Domains:** <list strong_domains or selected_interest_domains> > - **Technical level:** <technical_level, if present> > - **Tier:** <icp_tier, if present> > - **Mode last time:** <interview_mode> > > Would you like to: > - **"Use it"** — Keep this profile and move on to idea generation. > - **"Update"** — Run the interview again (full or short version) to refresh your profile. > - **"Browse topics"** — Pick new topics to explore, keeping the rest of your profile.

**Routing (existing profile):**

• User says "use it", "keep it", "yes", "that's fine" or similar → keep the existing profile, proceed to the next skill in the workflow. Do not overwrite `user_profile.md`.
• User says "update" → proceed to the **Opening Message** below (full interview options).
• User says "browse topics" → go to **Browse Path** below. Merge `selected_interest_domains` into the existing profile (preserve other fields).

**If no profile exists**, proceed directly to the Opening Message.

Opening Message

Before asking the first question, say exactly this:

> I'm going to ask you **10 questions** to understand your background, skills, and interests. This helps me find app ideas that genuinely fit you — not generic ideas, but ones where you have a real edge. > > Answer as freely as you'd like. There are no right or wrong answers. > > **Not sure where to start, or prefer not to answer questions?** No problem — pick one: > > - **"Browse topics"** — I'll show you real product domains people pay for. Pick 2–3 that interest you, and I'll generate ideas around those. > - **"Short version"** — Just 4 quick questions instead of 10. > - **"Skip"** — Jump straight to idea generation with generic recommendations.

**Routing:**

• User says "browse", "show me topics", "browse topics", "I have no idea", or similar → go to **Browse Path** below.
• User says "skip" → go to **Skip Path** below.
• User says "short version", "fast", "quick" or similar → go to **Fast Path** below.
• Otherwise → proceed with the **full interview**.

Skip Path

If the user wants to skip entirely, still ask the mandatory technical ability question:

> No problem — I'll skip the interview. Just one quick question I need to generate relevant ideas: > > **What's your technical ability?** (no-code / beginner / intermediate / expert)

If the user refuses to answer even this, default to "beginner" and note it.

Then write a minimal profile:

{
  "strong_domains": ["unknown"],
  "hidden_opportunities": [],
  "natural_problem_spaces": [],
  "distribution_advantages": [],
  "inner_circle_domains": [],
  "technical_level": "no-code | beginner | intermediate | expert",
  "fit_score_by_niche": {},
  "interview_summary": "User skipped the background interview. Technical level: <level>. Profile is generic — recommendations will not be personalized to domain fit or distribution advantages.",
  "interview_mode": "skipped",
  "all_questions_and_answers": []
}

Then proceed to the next skill in the workflow. Downstream skills will use default/neutral values when profile data is missing.

Browse Path (Topic Picker)

When the user has no idea what they want to build or prefers not to answer personal questions, present product domains they can browse and pick from. These domains are organized around core human desires (from `memory/extra-context/core-human-desires.md`) but presented as concrete, relatable product categories — not psychology theory.

How it works

1. Show **Batch 1** (5 domains). Each domain has a plain-language name, a one-line "why people pay for this", and 3 concrete product examples. 2. After each batch, show a clear action prompt (see "After each batch" below). 3. The user picks domains by number or name. Selections accumulate across batches. 4. After each batch, the user either:

• Says **"more"** (or similar) → show the next batch
• Says **"continue"** / **"let's go"** / **"generate"** (or similar) → proceed with selected domains

5. Show up to 4 batches (20 domains total). If they haven't picked after all batches, ask: "None of these? Tell me in your own words what kind of problems or people interest you." 6. Once the user says "continue" with 2–3 domains selected, ask one quick follow-up: "Last question — what's your technical level? (no-code / beginner / intermediate / expert)" 7. Write the profile and proceed.

After each batch

After presenting each batch and receiving the user's response, show this status prompt:

> **Your picks so far:** <list selected domains, or "none yet"> > > Pick **2–3 domains** total. Then: > - Say **"more"** to see the next batch of topics > - Say **"continue"** to start generating ideas with your selected topics

If the user already has 2–3 selections and says "more", show the next batch but remind them: "You've already picked <N> — feel free to swap any out, or say **continue** whenever you're ready."

If the user says "continue" with **fewer than 2** selections, prompt: "Pick at least one more — it helps me generate better ideas. Or say **more** to see another batch."

If the user says "continue" wit