The three layers
Every workspace is built from the same three layers, whatever the use case:
- Structure — a folder tree that mirrors how the work actually happens: Company/, Product/, Customers/ for a founder; Src/, Docs/, Prompts/ for a developer.
- Content — Markdown documents inside those folders: templates, checklists, briefs and worked examples you replace with your own material over time.
- Instructions — an AI Instructions/ folder that tells each tool how to navigate the rest: what lives where, which conventions to follow, what good output looks like.
Why folders beat prompts
A prompt is one message. A workspace is standing context. When your AI editor opens a ContextOS folder, it doesn't need to be told what your project is on every request — it reads the structure once and carries it through the whole session.
That's why the same model produces noticeably better output inside a workspace: it cites your actual documents instead of guessing, and it follows your conventions instead of inventing its own.
What happens when you open one
Extract the .zip into your project root and open the folder in your tool of choice. Claude Code reads AI Instructions/claude.md automatically. Cursor picks up .cursorrules. ChatGPT and Gemini take the instruction files as project context.
From the first message, the AI knows the map. Everything after that is just work.