interview-me Skill for Claude: Interview You Before Building
interview-me makes Claude ask you one question at a time, with its guess attached, until it knows what you actually want to build.
- Installs
- 61.6K
- on skills.sh
- Stars
- 112.6K
- on GitHub
- Updated
- Nov 2026
- last repo commit
This skill, from Addy Osmani's agent-skills collection, runs before any plan, spec or code exists. Claude states its best guess of what you want with a confidence number, then asks one question at a time, each with its own guess for you to react to. The SKILL.md triggers on an underspecified ask, or when you say things like 'interview me', 'grill me', 'are we sure?' or 'stress-test my thinking'.
It stops when Claude can predict your reaction to the next three questions it would ask. Then it writes back a short statement of intent and waits for an explicit yes. The skill needs a live person, so it says not to run it in CI, scheduled runs or autonomous loops.
Good for
- Anyone who types 'build me X' and gets the wrong thing back
- Product owners shaping a vague request before a spec exists
- Developers who want assumptions surfaced before code is written
Install interview-me
Claude Code
- 1
Copy this prompt.
Install the interview-me skill in Claude Code. Run this command: npx skills add https://github.com/addyosmani/agent-skills --skill interview-me Then check that it loaded and tell me how to use it.
- 2
Paste it into Claude Code. It shows the change and asks before running it.
Requires Node.js.
From its listing on skills.sh. Source.
What's in the skill
When it loads
Extracts what the user actually wants instead of what they think they should want. Achieves this through one-question-at-a-time interview until ~95% confidence about the underlying intent. Use when an ask is underspecified ("build me X" without "for whom" or "why now"), when the user explicitly invokes ("interview me", "grill me", "are we sure?", "stress-test my thinking"), or when you catch yourself silently filling in ambiguous requirements before any plan, spec, or code exists.
Sections in SKILL.md
- Overview
- When to Use
- Loading Constraints
- The Process
- Step 1: Hypothesize, with a confidence number
- Step 2: Ask one question at a time, each with a guess attached
- Step 3: Listen for "want vs. should want"
- Step 4: Restate intent in the user's own words
- Step 5: Confirm — explicit yes, not "whatever you think"
- Step 6: Terminal Turn & Stop Rule (CRITICAL)
- The 95% Confidence Stop
- Output
- Example
- Interaction with Other Skills
- Common Rationalizations
- Red Flags
Key facts
- Installs on skills.sh
- 61,638 as of 2026-11-02
- GitHub stars (repo)
- 112,555 as of 2026-11-02
- Last repo commit
- 2026-11-01
- License
- MIT
- skills.sh audits
- Gen Agent Trust Hub pass, Socket pass, Snyk pass
What it includes
One-sentence guess plus a confidence number
Claude writes what it thinks you want and a 0 to 100% confidence. Below about 70% it adds the reason, such as who it is for or what success looks like.
Step 1: HypothesizeQ: and GUESS: pairs, one at a time
Each question comes with Claude's own guess and reasoning, so you can react to a wrong guess instead of writing an answer from scratch.
Step 2: One question, one guessProbe answers that sound like best practice
When you give buzzwords or defer to convention, Claude asks what you would want if you did not have to justify it to anyone.
Step 3: Want vs should wantSix-line statement you correct line by line
Outcome, user, why now, success, constraint and out of scope. The out-of-scope line is required.
Step 4: Restate intent'Whatever you think' and 'sounds good' do not count
Claude re-asks with two concrete options or asks what you would refine, and loops until you say yes.
Step 5: Confirm with an explicit yesDeliver the statement, offer next steps, stop
After the yes, Claude hands back the statement, offers to save it to docs/intent/[topic].md or draft a spec, and ends its turn without starting work.
Write ·
Step 6: Terminal turn and stop rule
From the official docs
Questions
What does interview-me do?
It makes Claude interview you before it builds anything. Claude asks one question at a time, attaches its own guess to each, and keeps going until it can predict your reaction to the next three questions. The result is a confirmed statement of intent.
How do I install interview-me?
Use the install box above. The README gives npx skills add addyosmani/agent-skills --skill interview-me as the single-skill command. The README says a per-skill install copies only skills/<name>/, not the repo-level references/ folder.
When does it trigger?
On an underspecified ask such as 'build me a dashboard' with no who or why, or when you say 'interview me', 'grill me', 'are we sure?' or 'stress-test my thinking'. It does not apply to clear tasks like renaming a variable or fixing a typo.
Why does it stop after I say yes?
The skill ends Claude's turn on purpose after delivering the confirmed statement. It does not start tools or downstream work, so you choose the next step. Claude may offer to save the statement to docs/intent/[topic].md, and the SKILL.md says to save only if you confirm.
Does it work in CI or scheduled runs?
No. The SKILL.md says it needs a live, responsive user and must not be invoked in CI pipelines, scheduled runs, /loop or autonomous loops. In those cases Claude should flag the unclear ask as a blocker.
Does it need scripts or an API key?
The skill is a single SKILL.md of instructions with no bundled scripts, and its text names no API key.
What comes after it?
The SKILL.md names optional follow-ups from the same collection: idea-refine to generate variations, spec-driven-development to write the requirements down, and planning-and-task-breakdown after that. It places doubt-driven-development at the opposite end of the timeline, as a review of a finished plan.
Information
updated Nov 2, 2026
- Repository
- addyosmani/agent-skills
- Folder
skills/interview-me
Sources
Docs, issue trackers and discussions this page is built from.
Show all 2Hide
- interview-me SKILL.md · github.com
- addyosmani/agent-skills README · github.com