See what to fix first. Then put your plan to work.

Open the complete report prompt if you are starting from an existing business map. Paste or attach its actual text along with your approved context.

See a fictional example report (PDF). This illustrates the output format; it is not a client result or a promise of a fixed page count. Download the example with its evidence ledger, action plan and fictional inputs.

Your AI creates the report from the facts you provide. Review and save BUSINESS-COCKPIT-REPORT.md, EVIDENCE-LEDGER.md and ACTION-PLAN.md. If your assistant can create a PDF, ask it to export the reviewed report. Otherwise save the Markdown or print its HTML version. The initial kit email contains instructions, not your completed personal report.

For client discovery, use public or client-approved information, keep the prospect’s files separate and confirm assumptions before offering a scope or price.

• Current reality and goals. Understand where the business is, where the owner wants it to go, and what prevents progress.
• Customers, offers and proof. Check who the business serves, what it sells, and how it knows the offer works.
• Marketing, leads and sales. Map how interest turns into customers and where follow-up breaks down.
• Delivery and customer retention. Follow the work after a sale and identify delays, capacity limits and repeat business.
• Business economics and cash flow. Establish dated numbers without guessing profitability or lost revenue.
• Website, tools, data and AI. Check what is actually used, connected and working before recommending software.
• Team, decisions and support. Make responsibility, available time and implementation limits explicit.
• Priorities and the first useful change. Turn the audit into an owned, testable plan rather than a shopping list.

Use the guide in your kit, or copy the full business audit guide and say: “Run the full business audit with me, one question at a time.”

Use SHARED-BRAIN-SETUP.md in your free kit, open the shared-brain guide, or start with this prompt. It adapts to an existing setup or walks through what is missing. A shared file is a starting point; verify read-back and updates before calling anything synced.

Help me give my AI assistants one shared business brain.

Adapt to my device, situation, existing tools and your actual capabilities. Reuse my current names, roles and source files. Do not invent bots, connections or facts. Ask only for missing decisions, one short question at a time.

1. Check my assistants first.
Ask which assistants I want to use. For the three-assistant method, help me make sure Grokbot, Muse and Codex/Dot (or my chosen equivalents) are installed or available through their supported apps/websites and I am signed in. Use current official instructions; verify the exact product rather than guessing from a nickname. One existing assistant is enough to start; do not require all three for basic cockpit building. Never claim you can read another assistant's chats or memory without access.

2. Reuse or set up my tools.
Check whether I already have an Obsidian vault, working Graphify installation, private GitHub repository and Notion workspace. If something exists, preserve it. If I am starting from zero, guide me through the missing setup in plain language after I authorize it; use available tools only within that scope. Explain any paid requirement before I subscribe. Graphify maps approved files and their relationships; it is not itself a sync service. If my device or assistant cannot support a tool, provide a practical manual path and mark that setup incomplete.

3. Export each assistant's context.
Starting with the assistant that knows me best, create a Markdown export with Identity, Preferences, Active projects & goals, Key facts & decisions, and Relationships. If I have no saved context, interview me briefly and create a starter. Date changing information, distinguish reported from verified, and leave unknowns explicit. Export only information I approve for these recipients. Exclude secret values and confidential customer/client records. Keep different clients separate.

4. Merge into one source of truth.
Help me collect all exports through supported attachments or manual paste. Merge duplicates and flag conflicts. My latest confirmed decisions control; do not guess unresolved facts. Preserve richer existing instructions and files. If I already have a master location, use it. Otherwise propose a private GitHub repository for the master copy/history, a local Obsidian copy, Graphify for connections between notes, and Notion for a readable project dashboard. Confirm the location before creating it. Keep durable facts in LLM.md, active work in BRAIN.md, and the setup/update instructions in START-HERE.md. Record how each copy is updated and which one wins when copies differ. Do not create competing masters or overwrite local edits.

5. Connect only what is supported.
Check each assistant's available connectors or MCP servers and configure authorized access to the correct accounts. If a direct connection is unavailable, show me exactly which files to upload and how to refresh them. Clearly label manual snapshots, read-only access and any verified automatic updates. Never put API keys in notes, GitHub, Notion, graph inputs or chat. Reuse protected credential stores and verify account identity without printing secrets. A connector being installed does not prove it works.

6. Prove the shared brain works.
With my approval, save a harmless test fact and version/date in the master. Have each assistant read it back and identify the source/version it actually read. Change the fact and repeat using the documented update process; do not imply direct communication between assistants. Mark an assistant unverified if you cannot observe its result. Finish with a short checklist: saved locations, working access, manual update steps, verified sync directions, and blockers. Do not call it synced until the checks pass.

Then help me finish one useful task using my reviewed context. Leave me a simple habit: save approved decisions in the master, refresh the copies, and verify important updates. This is shared context, not training model weights. Do not send, spend, publish or delete without my authorization.