You have ten markdown files that capture who you are, how you work, and what matters to you. They are portable on purpose — but they do nothing sitting in your downloads folder. Here is how to wire them into the tools you already use, so every AI starts the conversation already knowing you.
You don’t need all of these. Pick the one that matches how you already work and add the rest later.
No setup, no servers. Upload your files once and every conversation in that project starts knowing you.
Update the files in the project whenever your work changes. Stale context is worse than none.
The universal method. Paste your context into the system prompt or custom instructions of any tool — ChatGPT, Gemini, Claude, anything.
<user_context> [paste identity.md] [paste communication-style.md] [paste any other files that fit the job] </user_context> You have context about me above. Use it to inform your responses — my role, my projects, my voice, my preferences. Don't reference it explicitly unless I ask. Just know me.
For tools with a character cap (ChatGPT custom instructions), use just identity.md and communication-style.md — the highest value in the smallest space.
Claude Code reads CLAUDE.md for context. Point it at your files and every session starts loaded.
## Personal Context My Context Kit lives in ~/context-kit/. Read the relevant files when you need to know who I am, how I work, my current projects, or my preferences. Start with identity.md if you only read one.
Or paste identity, communication-style, and preferences-and-constraints straight into CLAUDE.md if you’d rather not reference a folder.
Expose your files as an MCP resource and any MCP-compatible tool pulls what it needs on its own — no copy-paste, and updates are automatic.
Start local with the filesystem server. If you want it across devices, ask Claude Code to help you stand up a small remote MCP server.
For a system prompt, don’t paste all ten — irrelevant context dilutes the useful kind. Match the files to the work.
Describe how you actually work, not how you wish you worked. Your agents need ground truth.
When the interviewer drafts a file, the fixes you make are where the real signal is. A rubber-stamped file is a mediocre one.
A good context file is one page, not five. Agents do better with dense, high-signal context than sprawling documents.
Projects change and priorities shift. A kit that’s six months stale gives your agents confident, wrong context — redo the interview when your work moves.
Run the interview, download your ten files, and come back here to wire them in.
Open Context Kit