Skip to content

Guide · Agent skill

How to use the Interview Me agent skill: install and trigger it

Updated · written with AI from the project's own documentation and checked against our catalogue data

Interview Me is an agent skill from addyosmani/agent-skills that makes an AI agent stop and question you before it builds anything. It is useful when a request is vague, such as "build me a dashboard" or "make it faster", and you want the real goal pinned down first. After this guide you will know what the skill does, how to install it, how to trigger it, and when to leave it alone.

Interview Meaddyosmani/agent-skills · MIT

Quick start

  1. 1Download the ZIP. Get the interview-me ZIP from the directory page and unzip it on your machine.
  2. 2Place the skill folder. Move it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder for one project.
  3. 3Or upload in the apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the skill instead.
  4. 4Make a vague request. Ask for something underspecified, or say "interview me" explicitly.
  5. 5Answer one question at a time. Reply to each question and correct the agent's guesses where they are wrong.
  6. 6Confirm the restatement. Check each line of the intent summary and reply with an explicit yes.
  7. 7Choose the next step. After confirmation, pick whether to save the intent, write a spec, or plan tasks.

What an agent skill is and what Interview Me makes the agent do

An agent skill is a folder containing a SKILL.md file that gives an AI agent instructions for a specific kind of task. The agent loads the skill when a request matches its description.

Interview Me extracts what you actually want instead of what you think you should want. The agent writes down a one-sentence hypothesis with a confidence number, then asks one question at a time, attaching its own best guess to each. It keeps going until it can predict your reactions to its next few questions, which the skill treats as roughly 95% confidence about your underlying intent.

  • It asks about who the request is for, why now, what success looks like, the main constraint and what is out of scope.
  • It restates your intent in a short structured list and waits for an explicit yes before moving on.
  • It stops after you confirm and offers next steps, such as saving the intent to docs/intent/[topic].md or drafting a spec with spec-driven-development.

Install Interview Me by downloading the ZIP

Download the ZIP for the interview-me skill from the directory page and unzip it. Place the resulting skill folder in your personal skills directory, ~/.claude/skills/, or in a project's .claude/skills/ folder if you want it available only inside that project.

Alternatively, you can upload the skill in the Claude apps under Settings → Capabilities → Skills.

  • Personal install: ~/.claude/skills/ (available across your projects).
  • Project install: .claude/skills/ inside the project folder (available only there).
  • App install: Settings → Capabilities → Skills in the Claude apps.

Trigger the skill with natural requests

You do not need a special command. The skill's description lists the situations where it applies, so you can trigger it by saying what you want in plain language. Explicit phrases the skill recognises include "interview me", "grill me", "are we sure?" and "stress-test my thinking".

It also triggers when an ask is underspecified, such as "build me X" with no mention of who it is for or why now.

  • "Interview me before we build a reporting tool."
  • "Grill me on this idea before any code."
  • "Stress-test my thinking about moving the checkout flow."
  • "Build me a dashboard." (underspecified, so the agent may interview you)

Work through an interview

Expect the agent to open with a hypothesis and a confidence percentage, then ask a single question with a guess attached. Reply in your own words. If the guess is wrong, say so; the skill is designed so that visible wrong guesses surface your real assumptions.

When the agent is confident, it will restate your intent as Outcome, User, Why now, Success, Constraint and Out of scope. Check each line. Only an explicit yes counts as confirmation. Replies such as "whatever you think" or "sounds good" prompt a follow-up question instead.

  • Answer the question that is asked, even if it feels small.
  • Correct any line of the restatement rather than accepting it to be polite.
  • If you are unsure, say so and the agent will offer you two concrete options to choose from.

Know what the skill needs and its safety limits

The source lists no required dependencies. The skill runs no scripts and makes no network calls. The package consists of a single file, SKILL.md.

The skill needs a live, responsive user. The source says not to use it in non-interactive contexts such as CI pipelines, scheduled runs, /loop, or autonomous loops. In those settings, an underspecified ask should be flagged as a blocker rather than guessed at.

  • Requirements: none listed.
  • Runs scripts: no.
  • Uses network: no.
  • Files: SKILL.md only.

Tips for getting value from Interview Me

Use the skill before any plan, spec or code exists. The source notes that switching costs rise once building starts, and that users tend to rationalise a misfit into a good-enough result.

Skip it for simple, unambiguous tasks. Renaming a variable or fixing a typo gains nothing from an interview.

  • Good fit: vague goals, competing priorities such as simplicity vs. flexibility, and features you suspect you may have built wrong before.
  • Poor fit: pure questions like "how does X work?", and mechanical operations like file moves or formats.
  • If you notice the agent asking about "should" rather than "want", answer with what you would do if you did not have to justify it to anyone.

What people use it for

  • Turning a vague request for a dashboard into a clear statement of who uses it and what decision it supports.
  • Stress-testing a feature idea before anyone writes a spec.
  • Clarifying a performance goal like "make it faster" by agreeing on what success looks like.
  • Surfacing out-of-scope items early so a project does not drift.
  • Catching disagreement between you and a stakeholder's stated goal before work begins.

Frequently asked questions

Is the Interview Me skill free to use?

The source lists the skill under the MIT licence and does not mention any price. Check the licence file in the repository for the exact terms.

What do I need to run the interview-me skill?

The source lists no requirements and says the skill runs no scripts and makes no network calls. It does need a live user, so it should not be used in CI pipelines or scheduled runs.

Where do I install an agent skill like interview-me?

Unzip the downloaded skill folder into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder. In the Claude apps you can instead upload it under Settings → Capabilities → Skills.

How do I trigger the skill?

Use phrases such as "interview me", "grill me", "are we sure?" or "stress-test my thinking", or make an underspecified request like "build me X". The agent can also start the interview on its own when requirements are unclear.

When should I not use Interview Me?

Skip it for simple, unambiguous tasks like renaming a variable or fixing a typo, for pure information questions, and for autonomous or scheduled runs. The interview only adds delay in those cases.

How does it compare to spec-driven-development or idea-refine?

According to the source, Interview Me comes before those skills. idea-refine generates variations from an idea, spec-driven-development writes requirements down, and doubt-driven-development stress-tests a plan after it is drafted.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Interview Me suits anyone who starts with a vague idea and wants the agent to pin down the real need before building, or who often finds they built the wrong thing. It is the wrong choice for simple, unambiguous edits, quick factual questions, or unattended automation that needs no live user.

Download the skill