Claude website or Claude Desktop
This is the recommended Claude method. It installs the complete Skill, so Claude uses the canonical instructions and reference files in every chat rather than a single project.
First, turn on code execution. The Skills section does not appear without it.
On Free, Pro, and Max, open Settings, then Capabilities and enable Code execution and file creation. If the Skills section is missing later, this setting is the reason.
Install from a desktop browser. The mobile apps are not documented to support skill upload.
Step 1: download the Skill
Your browser saves a file named clean-language.zip. Do not unzip it. The Skill is prose instructions and reference material only. It contains no executable code and makes no network calls.
Step 2: upload it to Claude
On the Skills page:
- Click the + button, then + Create skill.
- Click Upload a skill.
- Select the downloaded
clean-language.zipfile. - Make sure Clean Language is switched on in your Skills list.
Test it in a new chat. Copy this prompt, paste it, then add your text:
Apply Clean Language to this draft:
Fallback: use a Claude Project instead
Use this only if your organization has disabled Skills or code execution. It applies to one project rather than every chat.
- In Claude, click Projects, then New Project.
- Name it Clean Language.
- Under Project knowledge, click + and upload the saved file.
- Set the project instructions to:
Use the Clean Language instructions in project knowledge for every response unless I explicitly tell you not to.
ChatGPT
Use this method when your ChatGPT account shows Plugins and Skills. Your workspace administrator may need to enable personal Skills.
-
The file saves as
clean-language-instructions.txt. - Open ChatGPT.
- In the left menu, click Plugins.
- At the top of the Plugin Directory, click Skills.
- Click Create.
- Click Create with chat.
- Type:
Create a skill named Clean Language using the instructions in the file I am about to upload. - Attach
clean-language-instructions.txtand send the message. - When ChatGPT shows the completed Skill, click Install.
Test it by starting a new chat and typing: Apply Clean Language to this message: [paste your text]
Follow these writing rules for this conversation.Gemini website or Gemini app
The easiest reusable method is a custom Gem. Create it once on the Gemini website, then use it from the website, mobile app, or supported Google Workspace side panel.
-
The file saves as
clean-language-instructions.txt. - On a computer, open gemini.google.com.
- Open the menu on the left.
- Click Gems. If you do not see it, click Settings and help, then Gems.
- Click New Gem.
- Name it Clean Language.
- In the instructions box, type:
Follow the Clean Language instructions in the attached knowledge file for every response. - Scroll to Knowledge.
- Click Add files, then Upload files.
- Select
clean-language-instructions.txt. - Click Save.
To use it later, open Gemini, click Gems, then click Clean Language.
Other AI systems
Clean Language also works with Microsoft Copilot, local models, API-based assistants, and other systems that accept custom instructions, system prompts, reusable knowledge, or uploaded files.
- Save the file or copy its contents.
- Add it to your AI tool as custom instructions, project knowledge, a system prompt, or an attached reference file.
- Tell the AI:
Use the Clean Language instructions for every response unless I explicitly tell you not to.
How to use Clean Language
You can paste rough notes or finished text. Use one of these requests:
Rewrite this using Clean Language:
[paste text]
Proofread this using Clean Language.
Fix mistakes but do not change my tone:
[paste text]
Audit this for robotic AI language,
filler, contrast framing, and dash overuse.
Do not rewrite it:
[paste text]
Write this with no em dashes or en dashes:
[paste notes]
For technical users
Developers can install the full Agent Skill from the GitHub repository for ChatGPT, Claude Code, Gemini CLI, and compatible Agent Skills clients.
Trust and assurance
- The skill contains no executable code and makes no network calls.
- Every rule and reference file is publicly reviewable in the GitHub repository.
- Each tagged release provides a SHA-256 checksum, so you can verify the download.
- Clean Language is licensed under CC BY-SA 4.0.
- Platform behaviour on ChatGPT, Claude, and Gemini is set by those providers and is outside this project's control.