// AI TOOLS · CLAUDETTE

Claudette

A free, open-source Claude Code skill — the /debuzz command in Adnan Akil's "nobuzz" project — that takes Claude's verbose, clickbait-style reply and pipes it through the Google Gemini CLI to rewrite it in plain, theatrics-free English. Surfaced on GitHub and Hacker News in August 2026.

Last verified · 2026-08-21 · by Moe Ameen

What Claudette is

Claudette is the informal name for "nobuzz," a small open-source Claude Code skill by Adnan Akil that installs as a slash command, /debuzz. Its one job is to fix a specific annoyance: Claude's habit of narrating ordinary engineering work like a TED talk — reframing a simple bug as a "load-bearing assumption," delivering numbered revelations, and noting solemnly that the third one is the most instructive. Claudette strips that stylistic layer off and hands you the plain-English version.

The design choice that makes it work is that it does not ask Claude to fix its own prose. Instead it captures Claude's previous reply, writes it to a temp file, and pipes it through the Google Gemini CLI with instructions to rewrite in a flat, direct register, then prints Gemini's output verbatim. Using a second model to translate is deliberate — the README's premise is that prompting Claude alone rarely removes the style for long. If Gemini fails (usually an auth problem) you see the real error instead of Claude re-buzzing its own translation.

It ships three modes for three audiences. Colleague (the default) keeps every file path and code block intact with zero theatrics; manager compresses to about a third the length as what-happened / why-it-matters / what's-next with no code; director cuts to three to five executive sentences of outcome, impact, and ask. You invoke it as `/debuzz [mode] [text]` — with no text it de-buzzes Claude's last reply — and it also triggers on natural phrases like "say that in normal english."

Two honest caveats. First, the name is a joke: the README explicitly says it is "definitely absolutely not called 'Claudette'." Second, it is a workaround, not a fix — it requires Claude Code plus the Gemini CLI (`npm install -g @google/gemini-cli`), authenticated, and it rewrites output after the fact rather than changing how Claude writes. It is MIT-licensed and free. Treat the specifics as an early snapshot and confirm them in the repo before you depend on a detail.

What you can make with it

  • A plain-English rewrite of Claude's last reply, with every file path and code block preserved (colleague mode)
  • A manager-length summary of a long technical answer — what happened, why it matters, what's next, about a third the length
  • A three-to-five-sentence executive readout of outcome, impact, and ask (director mode)
  • A de-buzzed version of any text you paste in, not just Claude's previous message
  • A quick "say that in normal english" rephrase triggered by natural language rather than a command
  • A stripped-down changelog or standup note from a verbose engineering explanation

How Kompozy turns Claudette output into content

Claudette treats AI-tell as something you remove after the fact: Claude writes in its theatrical register, then a second model (Gemini) rewrites the result flatter. That is the right shape for one message in your terminal. It is the wrong shape for content you publish, because you would be running every caption, script, and blog draft through a second tool, one at a time, with no memory of your actual voice. [Kompozy](/) closes the same gap from the other end — it governs voice at the moment the text is generated, so there is nothing to de-buzz later.

Inside Kompozy, a [Persona Brief](/glossary/persona-brief) plus a banned-word filter define your register once and apply it to every generation: captioned [Persona Shorts](/glossary/persona-shorts) and avatar video, brand-exact [Carousels](/glossary/hyperframes), Photo Posts and Quote Graphics, Blog Articles, and Email Newsletters — 18 formats from one source, all in the same voice, no clickbait framing to strip out because it was never written in. Each asset clears a per-post review gate, then [Autopilot](/glossary/autopilot) schedules and publishes the batch across the eight social platforms plus blog and email. So keep Claudette for making your Claude Code sessions readable; reach for Kompozy when the output is content that has to sound like your brand and land on every platform. If you also rewrite already-generated copy, an AI humanizer like [Ryne AI](/ai-tools/ryne-ai) works on finished text the way Claudette works on a terminal reply.

  1. Use Claudette in Claude Code — /debuzz — to keep your own working notes, standups, and explanations in plain English.
  2. When you have a de-buzzed idea, angle, or rough script worth publishing, drop that text into Kompozy as a source.
  3. Set a Persona Brief and banned-word filter once so every generated asset carries your voice with no AI-tell to remove.
  4. Generate a full batch — Persona Shorts, carousels, quote graphics, a blog, and a newsletter — and review each in the per-post queue.
  5. Let Autopilot schedule and publish the set across the eight social platforms plus blog and email.

Frequently asked questions

What is Claudette?

Claudette is the informal name for "nobuzz," an open-source Claude Code skill by Adnan Akil that installs as the /debuzz slash command. It takes Claude's verbose, clickbait-style reply and pipes it through the Google Gemini CLI to rewrite it in plain, direct English. It surfaced on GitHub and Hacker News in August 2026 and is MIT-licensed and free.

Is it really called Claudette?

Not officially. The README jokes that it is "definitely absolutely not called 'Claudette'," so the real project name is nobuzz and the command is /debuzz. "Claudette" is the nickname people use for it, including in the Hacker News discussion.

How does Claudette de-buzz Claude's writing?

It does not ask Claude to fix its own prose. It captures Claude's previous reply, writes it to a temp file, and pipes it through the Gemini CLI with instructions to rewrite in a flat register, then prints Gemini's output. Using a different model is intentional — the premise is that prompting Claude alone rarely removes the style reliably. It offers colleague, manager, and director modes for different audiences.

What do I need to run Claudette?

Claude Code plus the Google Gemini CLI (npm install -g @google/gemini-cli), authenticated with a Gemini API key or CLI login. You install the skill by cloning the repo and copying the debuzz folder into ~/.claude/skills/. It is free, but the Gemini dependency is a second tool and potentially a second bill.

How is Claudette different from using Kompozy to fix AI-tell?

Claudette removes clickbait style from one Claude reply after it is written, in your terminal. Kompozy prevents it at the source for content you publish — a Persona Brief and banned-word filter govern every generation across 18 formats, so nothing needs a second model to de-buzz it, and Kompozy then schedules and publishes across the eight social platforms plus blog and email.

Related tools

  • Ryne AIAI humanizer that rewrites AI-generated text to read as human and slip past AI detectors.
  • Claude TagAnthropic's always-on Claude teammate that lives in Slack, learns from your channels, and works tasks in-thread.
  • Gemini AppGoogle's consumer AI assistant — chat, voice conversations with live camera and screen sharing, image generation, and deep research on Android, iOS, web, and desktop. It crossed one billion monthly active users in August 2026.
  • AI Creative Director with ClaudeA workflow that configures Claude as a creative director — it takes a rough idea through brief, concept, and finished per-platform drafts.
  • ChatGPTOpenAI's AI assistant for writing, ideas, and images — which, since late July 2026, refuses direct requests to write "in the style of" a specific named author.

← All AI tools · Get started →