interview-me

Extracts underlying user intent through one-question-at-a-time interviews before planning or coding.

1|Updated Mar 2, 2025
One-click install
npx skills add https://github.com/marjorg/setup --skill interview-me-marjorg
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: interview-me
Source: https://github.com/marjorg/setup/tree/main/home/.agents/skills/interview-me
Command: npx skills add https://github.com/marjorg/setup --skill interview-me-marjorg

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Users often ask for what they think they should want rather than what they actually need, and building on those unstated assumptions locks in the wrong solution. This Skill closes the gap between the stated ask and the real intent before any plan, spec, or code exists, when changing direction is still free. ## Core Features & Use Cases - Hypothesis-Driven Questioning: States a one-sentence hypothesis with an explicit confidence number, then asks one focused question at a time with a guess attached so the user can react instead of generating answers from scratch. - Want vs. Should-Want Detection: Identifies convention-driven or sophistication-signaling answers ("scalable", "best practice") and probes what the user would want if they didn't have to justify it. - Confirmed Intent Restatement: Produces a structured restatement (Outcome, User, Why now, Success, Constraint, Out of scope) gated on an explicit yes, with a 95% confidence stop condition. - Use Case: A user says "build me a dashboard for our metrics." Instead of proposing chart libraries, the Skill interviews them and discovers the actual need is a personal experiment tracking list, avoiding building the wrong artifact entirely. ## Quick Start Ask the agent to interview you about your request before it starts planning or writing any code.

Frequently Asked Questions about interview-me

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I clarify vague feature requests before writing code?

State a one-sentence hypothesis with a confidence number, then ask one focused question at a time with your best guess attached. Continue until you can predict the user's reactions, then restate the intent covering outcome, user, success criteria, and constraints for explicit confirmation.

How to handle users who say build me X without details?

Treat conventional requests like "build me a dashboard" as underspecified and interview before planning. Attach guesses to each question so users react rather than generate answers, and watch for answers that pattern-match best practices instead of stating real needs.

When should I not use a requirements interview approach?

Skip it for unambiguous self-contained tasks like renames or typo fixes, pure information requests, mechanical operations, or when the user explicitly prioritizes speed. It also requires a live responsive user, so avoid it in CI pipelines or autonomous runs.

Why does one question at a time work better than a question list?

Batched questions encourage skim-reading and surface answers, and later questions often depend on earlier answers. Single questions with attached guesses let users react quickly, expose your assumptions, and keep the framing correct as the interview progresses.

What counts as confirmed user intent versus a hollow yes?

Confirmation requires an explicit yes to a concrete restatement covering outcome, user, why now, success, constraint, and out of scope. Responses like "whatever you think" or "sounds good" are delegation or ambiguity, not confirmation, and require re-asking with concrete options.