Open-source · local-first · MIT

Walk into every round already prepared.

Most tools answer one question at a time. greenroom prepares the whole package from your real material — company and interviewer research, your experience turned into reusable cards, scripts you can say out loud, a realistic mock — all in one Markdown folder on your machine.

候场 — 上台之前,把每一句话备好、对好。简历、岗位调研、经历库、逐字稿、模拟面试,全部活在你电脑上的一个 Markdown 文件夹里,本地渲染,零上传。

Demo runs entirely in your browser on fictional data — nothing is uploaded.

8Claude skills
15industries
155role playbooks
1Markdown folder
0bytes uploaded
Verbatim, not outlines

Scripts you can actually say out loud.

Most tools hand you a bullet list or an obviously AI-written paragraph you’d never speak. greenroom writes spoken-word scripts to a real style standard — stance, anti-cliché rules, on-the-spot readable formatting — and matches your voice from a writing sample. The output is what you’ll actually say in the room, in 中文 or English.

read it → sounds like you → not like a bot
How it works

One sentence to Claude, or one screen with no Claude at all.

Tell the greenroom skill your resume, target company and the JD — it runs the pipeline end to end. No Claude? The console’s “Add role” screen generates the workspace from a dropped resume + pasted JD.

1

Job intel

Company & interviewer research, JD matched line by line, next-round questions predicted.

2

Experience bank

Real experience into sourced cards — facts once, angle per role.

3

Industry brief

Landscape, role playbook, the concepts you must be able to explain.

4

Verbatim script

Spoken-word scripts you can read aloud — not an AI-flavored outline.

5

Mock interview

An interviewer persona built from the real research, with a coach report.

6

Debrief

Reconstruct the round, record what you actually said, constrain the next one.

+

Live copilot

In-round prompting fed by the same workspace, a floating window under your camera.

Roles encyclopedia

Function × industry: 155 playbooks across 15 industries, esports to silicon.

Why it’s different

Built like a product you’d trust with your most sensitive data.

Local-first by design

Your workspace is one folder of Markdown on your machine. The console parses it in the browser. Resume, salary, reasons for leaving — none of it leaves your computer.

Verbatim, not outlines

Scripts written to a spoken standard — stance, anti-AI-cliché rules, readable on-the-spot formatting. Give it a style sample and it matches your voice.

Consistency across rounds

The claim ledger makes resume, three interviews and the background check all hear the same version of every fact.

Your engine, swappable

Claude skills today; the workspace contract is open. Any OpenAI-compatible model powers the reference backend — your key, your model.

One file, zero deps

The whole console is a single HTML file — no build, no framework. Open it, drag your folder in, done.

Honest about screen-share

No fake “undetectable” claims — macOS 15.4+ killed window-hiding for everyone. The only safe prompt is a second device, and we tell you so.

Get started

Install the skills, or just open the console.

# Claude Code — plugin, versioned updates /plugin marketplace add YunyueLi/greenroom /plugin install greenroom@greenroom # or any agent that supports skills npx skills add YunyueLi/greenroom # run the console with your own folder (zero deps) python3 serve.py ~/my-greenroom

No setup at all? Open the live demo → — it runs the full console on a fictional candidate, right here in your browser.