Guide · Agent skill
How to use the Product Design User Context agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The user-context skill is a small agent skill from the XiaomiMiMo/MiMo-Code project, part of its Product Design workflows. It lets an agent keep a saved notebook of the product URLs, Figma files, screenshots, codebase paths and design tokens you reuse, and read that notebook at the start of later design tasks. This guide explains what the skill does, how to install it, how to trigger it, and the limits you should know before relying on it.
Product Design User ContextXiaomiMiMo/MiMo-Code · MIT Download the skill Quick start
- 1Download the ZIP. Download the user-context skill ZIP from this page.
- 2Unzip into a skills folder. Unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder.
- 3Or upload in the Claude apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the skill.
- 4Check requirements. Confirm Python 3 is installed and that you have local shell access and a writable CODEX_HOME or ~/.codex directory.
- 5Ask the agent to set up Product Design. Make a plain request such as "set up Product Design" and the agent will explain what it can remember.
- 6Provide your sources. Give product URLs, Figma files, screenshots, codebase paths or tokens. Do not include secrets.
- 7Verify saved context. Ask "what does Product Design remember?" to see the saved entries.
What is an agent skill, and what does user-context do?
An agent skill is a folder of instructions and optional scripts that an AI agent loads when a request matches the skill's description. The user-context skill covers saving and reading Product Design context: the product and design sources you use often.
Once saved, the agent uses this context as default grounding. Ideation, prototypes, audits, clones and critiques should match your saved product context unless you ask for something different. When visual work needs it, the agent can also include saved screenshots, reference images, tokens and component references.
- Saves product URLs, Figma files, screenshots and reference images
- Saves codebase paths, Storybook and component docs, design tokens and theme sources
- Saves brand, logo, icon and illustration sources
- Saves preferences such as browser, capture tools and share targets
- Lets you ask what Product Design currently remembers
How do you install the user-context skill?
You can install the skill in two ways. Download the ZIP from this page and unzip it into your personal skills folder at ~/.claude/skills/, or into a project's .claude/skills/ folder if you want it limited to one project.
Alternatively, in the Claude apps, open Settings, go to Capabilities, and upload the skill under Skills.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into the project's .claude/skills/ folder
- Claude apps: Settings → Capabilities → Skills, then upload
What does the skill need to run?
The skill needs local shell access and Python 3, because its scripts run with python3. It also needs a writable CODEX_HOME directory, or a ~/.codex directory, to hold the state file. Without these, saved context cannot persist between conversations.
The skill checks this before offering onboarding. It confirms shell access, that CODEX_HOME can be resolved, and that the state folder exists or can be created and is writable. If a check fails, the agent should say that context can be used in the current conversation only.
- Python 3 (scripts run with python3)
- Local shell access
- A writable CODEX_HOME or ~/.codex directory for the state file
- The skill does not use the network
How do you trigger the skill with natural requests?
You do not need a special command. Ask in plain language. The skill's description matches requests about setting up Product Design, getting started, onboarding, saving product or design sources, seeing what Product Design remembers, updating saved context, or remembering a preference.
For a first run, ask the agent to set up Product Design. It will explain what can be remembered and ask for useful sources before writing anything. Setup does not inspect your workspace, install dependencies, or generate images.
- "Set up Product Design for my product."
- "Save this Figma file as a product design source."
- "What does Product Design remember about my product?"
- "Remember that I prefer Chrome for product captures."
- "Update my saved Storybook link."
How does saving and reading context work?
Saved context lives in a file named user-context.md inside $CODEX_HOME/state/plugins/product-design/, with screenshots and reference images in an assets/ folder beside it. When you give screenshots to save, the agent copies them into assets/ with a descriptive filename, such as assets/checkout-confirmation-screen.png, and links them from the entry.
When a Product Design task needs saved context, the agent runs python3 scripts/user_context_preflight.py and uses the returned entries as its starting point. It does not open every saved reference, only the ones the current task needs. To create the file during setup, the agent runs python3 scripts/init_user_context.py and then adds your references to it.
- State file: $CODEX_HOME/state/plugins/product-design/user-context.md
- Images: $CODEX_HOME/state/plugins/product-design/assets/
- Descriptive image names make later runs easier to follow
python3 scripts/user_context_preflight.pyWhat safety rules should you know about?
The skill runs scripts, and the scripts write plain-text files. Do not store secrets, credentials, API keys, private tokens or copied customer data in saved context. The skill's own rules forbid saving these, but the files are not encrypted, so anything you put in them could be exposed to anyone who can read that folder.
The skill does not use the network, but it runs local Python scripts, so review the scripts in the scripts/ folder if you need to confirm what they do before installing.
- Never save API keys, tokens, passwords or customer data
- Saved files are plain text and not encrypted
- Keep saved context curated: a few high-value references beat a long dump of URLs
Tips for getting the most from user-context
Save the sources you reuse most: your main product URL, the Figma file your team designs in, your token file and your Storybook. Ask the agent what it remembers after a few sessions to check that entries are still current.
Use clear image names and update entries when a product changes. If the agent says context cannot be saved, check the shell access, CODEX_HOME and folder permissions named in the checks above.
- Start with a handful of sources, not everything
- Check saved context periodically with a "what do you remember" request
- Name screenshots by what they show
What people use it for
- A designer saves the Figma file and brand logo sources so that prototypes use the right components and colours.
- A team saves its Storybook URL and token file so that audits and critiques reference the real design system.
- A designer saves a key product flow URL and a screenshot of the settings modal so later ideation starts from the current UI.
- A user records a preference such as a preferred browser for captures so future runs use it.
Frequently asked questions
How do I install the user-context agent skill?
Download the ZIP from this page and unzip it 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.
What does the user-context skill require?
It requires Python 3, since its scripts run with python3, and local shell access. It also needs a writable CODEX_HOME or ~/.codex directory for the state file. The skill does not use the network.
Does the user-context skill run scripts?
Yes. It runs scripts/init_user_context.py during setup and scripts/user_context_preflight.py when a Product Design workflow needs saved context. The scripts write plain-text files.
Can I store API keys or passwords with this skill?
No. The skill's rules forbid saving secrets, credentials, API keys, private tokens or customer data. Its files are plain text and not encrypted, so keep them out of the saved context.
Why doesn't the agent remember my context between conversations?
Saved context needs local shell access and a writable state folder. If any check fails, the agent should say that your references can be used only in the current conversation and cannot be saved for future ones.
Is user-context free and what is its licence?
The project is released under the MIT licence. This source does not state a price, so check the project page for any cost details.
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
Use user-context if you are a designer or team who repeatedly works from the same Figma files, brand assets, product links and design system, and you can run local scripts with Python 3. Pick something else if you need to store secrets or credentials, or if you work in an environment without shell access or a writable state folder, since saved context will not persist there.
Download the skill