How to use your brand.OS file pack

When your file pack arrives by email, you'll have:

1. A zip of .txt files: around 10 for a single language, with each extra language adding 8 more. These are the brand DNA files your AI reads as its authoritative source of truth.

2. A custom-instructions block: one paste-in block of text that tells your AI how to use the pack. You'll find it in the delivery email and in the setup guide, ready to copy.

3. A guided setup link: an interactive walkthrough that stands the pack up step by step, for whichever platform you pick. This page is its detailed companion.

We recommend Claude. Reasons:

• Claude Projects loads every uploaded file fully into context, every conversation. The file pack relies on cross-file references, and this matters.
• Claude's instruction-following tends to be tighter for complex multi-rule systems.
• The framework was designed and stress-tested primarily on Claude.

ChatGPT works, with caveats:

• ChatGPT uses retrieval over knowledge files. For larger or many files, only chunks may be retrieved per query.
• GPT models can be more variable in adhering to strict system rules.
• Output reliability on complex analytical work is somewhat lower.

The others work too. The files are plain text, so the pack runs anywhere you can attach reference files and set instructions:

• Gemini (Gem): a free Gem holds your files and instructions. Like ChatGPT, it retrieves over files rather than loading them whole.
• Grok (Workspace): a Workspace holds your files and custom instructions. Keep the files as plain text; setup may need a paid Grok plan.
• Agent orchestrators (Hermes, OpenClaw, and similar): point the agent at the files as its context and paste the instructions in as its system prompt.

For routine content and most analytical work, any of these is fine. For high-stakes outputs or multi-locale brands, prefer Claude.

1. Sign in to claude.ai
2. Create a new Project (named after your brand). Use this Project exclusively for brand reasoning. Don't mix with general chat.
3. Upload all numbered .txt files from your zip to the Project's sources panel.
4. Open the Project's Custom Instructions field.
5. Paste the custom-instructions block (from the delivery email), exactly as provided. Don't modify.
6. Save.
7. Start a new conversation. Type: system check
8. You should see: brand name, system version, files loaded, ready confirmation.
9. Run a few test prompts to verify behavior.

If system check shows wrong values or no "Ready" confirmation, return to step 3.

One project, shared once, becomes the single source of truth for the whole department. You maintain it; everyone else just uses it. Update it once and the new version reaches everyone, with no re-sending and no drift between people's copies.

Sharing requires a Claude Team or Enterprise plan. On Pro or Free you can't share a project, so each person would set up their own copy and they'd drift apart over time. If you're on Pro, upgrade to Team first, then share.

To share (Team / Enterprise):

1. Open the brand project in claude.ai.
2. Click Share, at the top-right of the project, just right of the project name.
3. Add your colleagues by email, or share with your whole org.
4. Set most people to Can use: they chat with the brand but can't change the files.
5. Keep Can edit for yourself (and anyone who owns the brand), so the canonical files stay controlled.

Once shared, the project syncs across everyone's devices and apps, including phones. No one repeats the setup steps. New team members just open the shared project and start working on-brand.

1. Go to chatgpt.com (a paid plan is required to build a Custom GPT)
2. Create a new Custom GPT
3. In the Knowledge section, upload all numbered .txt files from your zip
4. In the Instructions field, paste the custom-instructions block, exactly as provided
5. Save the GPT (private or shared, your choice)
6. Start a conversation. Type: system check
7. Verify the response. Run a few test prompts.

1. Go to gemini.google.com (a free plan works)
2. Open Gems and create a New Gem (named after your brand)
3. In the Knowledge section, upload all numbered .txt files from your zip
4. In the Instructions field, paste the custom-instructions block, exactly as provided
5. Save the Gem
6. Start a conversation. Type: system check
7. Verify the response. Run a few test prompts.

1. Go to grok.com (setup may require a paid Grok plan)
2. Create a New Workspace (named after your brand)
3. In Files, upload all numbered .txt files from your zip. Keep them as plain text.
4. In the Custom instructions field, paste the custom-instructions block, exactly as provided
5. Save the Workspace
6. Start a conversation. Type: system check
7. Verify the response. Run a few test prompts.

File-based agents don't have a hosted project to upload to, so the steps are generic. The pack is just text, which is exactly what these tools expect.

1. Put all numbered .txt files where your agent reads its context (a project folder, a context directory, or wherever it loads reference files).
2. Paste the custom-instructions block into the agent's system prompt, exactly as provided.
3. Start a run and type: system check. Verify the response, then run a few test prompts.

Tools differ in how they load files and set a system prompt. If yours needs specific wiring, reply to your delivery email and we'll tailor the steps for it.

Your brand reasoning system detects mode automatically from each prompt.

Content mode: for deliverables.

• Generating new content: "Write an Instagram post about our new community."
• Improving drafts: "Make this email more professional: [paste draft]"

You don't need to flag that you're submitting a draft. Just paste it. The system uses it as starting material and produces the brand-aligned version.

Analytical mode: for thinking and evaluation.

• Open questions: "Should we expand to Spain?" "Why is vertical integration cheaper?"
• Evaluating specific proposals: "I'm thinking of doing custom concrete builds for high-end clients."

For specific proposals, the system produces a structured trade-off analysis: where it aligns with your brand, where it creates tension, trade-offs to consider, paths forward.

You don't specify mode. The system reads what you're asking and routes accordingly.

Just type these in plain chat. They work the same way on every platform (Claude, ChatGPT, Gemini, Grok, and file-based orchestrators):

The file pack is a brand-grounded drafting and reasoning partner, not a final authority. For some scenarios, human review remains essential:

• Legal communications and contracts
• Regulatory filings or financial disclosures
• High-stakes strategic decisions (use the system to reason; commit through your usual governance)
• Sensitive personnel communications
• Public crisis responses

Use the system to think and draft. Use human judgment to commit and send.