prep-interview

name	prep-interview
description	Build an interview prep brief — 8–12 likely questions split across the loop's rounds, profile-grounded STAR talking points per question, a prep checklist, and diplomatic questions to ask back. Reads companies/<slug>.json from research-company if present; otherwise runs a thin in-line sweep. Forked context.
when_to_use	Use before any interview round (recruiter screen, technical, hiring manager, panel, final / onsite). Compose with /research-company first for the richest output.
user-invocable	true
allowed-tools	Read, Write, WebSearch, WebFetch
context	fork

prep-interview

Walk into the room ready. Hard rules: ../_shared/RULES.md. Question quality bar: reference/prep-rubric.md. Verb sourcing per tailor-resume/reference/action-verbs.md.

Inputs

1. company + role (required).
2. Profile — profile/master-profile.md (Notes / voice for tone; experience for STAR stories).
3. Optional company brief — companies/<slug>.json from a prior /research-company run. If present, use it heavily; if absent, run a thin sweep here (≠ a full research pass).
4. Optional round — --round screen | technical | hiring-manager | panel | final. If omitted, generate questions across the whole likely loop.

Method

0. Region pack. Read knowledge/regions/<code>.md — US: behavioral often STAR-formal; IN: more credential-led at large MNCs. Note tone adjustments.

1. Hydrate the company. If companies/<slug>.json exists, harvest:

   • jobSummary.dayToDayResponsibilities → likely "what would your first 30/60/90 look like" / "walk me through how you'd ..." Qs.
   • news.recentNews → 1–2 "saw the X announcement — how would you think about Y" Qs.
   • red_flags → honest concerns the user should diplomatically raise back.

   If absent, run a thin sweep: 3–5 WebSearch hits about the role + product + any recent news; no Reddit / Blind pass (that's research-company's job, not this one).

2. Identify the loop. Map target.tracks to the canonical rounds from ../mock-interview/reference/report-schema.md:

   • software → coding · system-design · behavioral
   • product → product-sense · execution · strategy · behavioral
   • design → portfolio-review · whiteboard/app-critique · behavioral
   • data → coding/SQL · ML/stats · case · behavioral
   • quant → math-stats · research · coding · behavioral
   • finance → technical · case · fit
   • marketing → case · craft · behavioral
   • founders-office → case · ops · behavioral
   • sales → discovery/role-play · pitch/demo · behavioral
   • operations → case · execution · behavioral
   • customer-success → scenario/role-play · case · behavioral
   • content → portfolio/clips-review · writing-exercise · behavioral
   • hr → case · craft · behavioral

   Cap rounds to those realistic for target.seniority (no system-design for entry-level SWE; behavioral always present).

3. Generate 8–12 questions total, split across rounds. Mix per reference/prep-rubric.md:

   • Generic to the role/round (the predictable spine).
   • Company-specific (drawn from the brief).
   • Profile-targeted (a hiring manager will ask about the biggest claim on the resume — prep for it).

4. For each question, emit 1–3 STAR talking-point bullets grounded in the profile — no invented stories. STAR = Situation · Task · Action · Result. Keep bullets fragmentary (the user will fill them in conversationally).

5. Prep checklist — 4–8 items. Things to read / refresh / practice / prepare to ask back. Include honest concerns from red_flags for the user to probe diplomatically.

6. Questions to ask back — 3–6, drawn from dayToDayResponsibilities, news, and red_flags. Each one frames a real concern as a constructive question.

Output

interviews/<company-slug>-<role-slug>.prep.md

# Prep — <Company> · <Role>
> as of <date> · source: companies/<slug>.json | thin sweep

## What to know about the company
- 5 bullets, each with a source link

## Loop & questions

### Round 1 — <name>
**Q1.** ...
  - STAR: ...

**Q2.** ...
  - STAR: ...

### Round 2 — <name>
...

## Prep checklist
- [ ] ...

## Questions to ask back
- (honest concern, diplomatic framing): "..."
- ...

Stdout

3-line summary — # questions across # rounds · 1 most important Q · suggested next (/mock-interview with this prep + role).

Anti-patterns

• Generic SWE Qs in a PM loop (or vice-versa).
• Inventing project stories the profile doesn't show.
• "Tell me about yourself" filler stacked on multiple rounds (one is fine if a screen leads).
• Quoting context.career_goal / context.additional_info.
• Cookie-cutter "Tell me about a time you failed" without a profile-grounded angle.
• Asking-back questions that are gripes in disguise — keep them constructive.
• Emoji.

Rubric

See reference/prep-rubric.md. Plus:

• 8–12 Qs total; balanced across the loop's rounds.
• ≥ 3 Qs are company-specific (cite source from brief or sweep).
• Every STAR bullet cites a profile fact (experience[i], project, skill).
• Checklist has ≥ 1 item per round.
• Questions-to-ask-back includes ≥ 1 diplomatic raise of a red_flag from the brief (when one exists).

Next steps

Prep brief ready. Next: /mock-interview to practice it.