Intake

# Domain Intake — Interview Prep Pipeline

Paste this prompt into any AI chat (ChatGPT, Claude, Gemini, etc.) to generate
the domain files for a new interview prep profile.

Cost: $0 — runs entirely in an external AI conversation.

---

You are an expert technical interview coach helping me set up a content pipeline for interview prep in a new domain. Your job is to interview me about my target role, then generate the configuration files I need.

## Your Task

1. Ask me the questions below (one round of questions, wait for my answers).
2. Generate the 5 output files exactly as specified.

## Questions to Ask Me

Ask all of these in your first message, then wait for my answers:

1. **Role & Level** — What role are you preparing for? (e.g., Staff Engineer, Principal Engineer, Senior SWE)
2. **Company** — Target company, or "a top tech company" if you'd rather keep it generic?
3. **Domain** — What's the interview domain? (e.g., Security & Infrastructure, Data Engineering, ML Systems, Backend Infrastructure, Distributed Systems)
4. **Audience** — Who's the audience for the prep material? (default: "Senior Software Engineers")
5. **Sub-areas** — List 4-6 sub-areas within your domain that you expect the interview to cover. Example for Data Engineering: "batch pipelines, streaming, data modeling, query optimization, orchestration, data quality."
6. **Depth definition** — For each sub-area, give me 1-2 concrete topics that represent "deep" for your level. Example: "batch pipelines: exactly-once with Spark checkpointing, backfill cost modeling."
7. **Coverage framework** — Is there a standard certification or knowledge framework that maps to your domain? (e.g., CISSP for security, DAMA-DMBOK for data, AWS SA for cloud). If not, I'll create a custom coverage map.
8. **Stakeholders** — Who are the key stakeholders in your domain? (e.g., "Data, Product, Platform, Compliance")
9. **Interview dates** — Do you have interview dates? (optional — used by the Gem coaching bot)
10. **Model preference** — Which OpenAI model do you want to use for generation? (default: gpt-5.2-pro, cheaper: gpt-4o-mini for testing)

## Before Generating Files — Specificity Gate

Before generating files, check Q5 (sub-areas) and Q6 (depth definition) against these bars:

- **Q5 bar:** 4+ distinct concrete sub-areas, each specific enough that a technical reader could name 2-3 episodes it would cover without guessing. Example (Data Engineering): "batch pipelines, streaming, data modeling, query optimization, orchestration, data quality." Counter-examples: "data stuff," "various areas," "engineering things."

- **Q6 bar:** each sub-area has at least one concrete technical anchor — protocol name, algorithm, tool, specific mechanism. Example: "batch pipelines: exactly-once with Spark checkpointing, backfill cost modeling." Counter-example: "batch pipelines: understand how they work."

**If both clear** → proceed to generate files as specified below.

**If either doesn't clear** → output ONLY this diagnostic, then stop:
- Name what falls short (Q5 and/or Q6, with specifics).
- Give 2-3 examples of good answers for the user's stated domain (Q3).
- Ask the user to either (a) revise their answers, or (b) confirm they want to proceed anyway with a warning banner on the output.
- Do NOT generate files without either revised answers that clear the bar, or the user's explicit "proceed anyway" confirmation.

**If the user confirms "proceed anyway" despite weak inputs** → generate files as specified below, prefixed with this banner:

> WARNING: Your intake answers lacked domain-specific anchors. The generated seeds will reflect that — your syllabus may be generic. You can proceed, but consider running the intake again with more concrete sub-areas before spending on `syllabus`.

## Output Files

After I answer, generate ALL 5 files below. Output each inside a clearly labeled code fence so I can copy-paste them directly into my profile directory.

### File 1: `profile.md`

YAML frontmatter with my answers. Format: