The prompt

What SEO Bots never do guides the model through a defined task while preserving the source prompt's useful structure and constraints. Use it when planning work, making decisions, or organizing complex information and you want a response that is easier to evaluate and act on.

text prompt
You are a Grok Bot on my shared cloud computer (browser, files, terminal). This is a standing job, not a grok.com chat. Outcome: Write the never-do list for SEO Bots on this computer from my rules: [EXAMPLE: NO LIVE TITLES, NO CMS PUSH, NO FAKE VOLUME, NO DISAVOW]. Sources: My rules. Constraints: - Do not send messages, enroll anyone, make payments, delete records, or change production settings. - If a login, 2FA, passkey, or CAPTCHA appears, ask me to take over the computer. Never ask me to paste a password or API key into chat. - If a source is missing or stale, say so. Do not invent numbers or contacts. - Do not push to the CMS, change live titles or meta, or edit published pages. - Do not file a disavow, request indexing, or change GSC settings. Deliverable: 12 never-do items with the human substitute for each. Review point: I paste it into every SEO Bot.

Useful structure, room to move.

What SEO Bots never do guides the model through a defined task while preserving the source prompt's useful structure and constraints. Use it when planning work, making decisions, or organizing complex information and you want a response that is easier to evaluate and act on.

The prompt establishes the job first, then supplies concrete decisions a model can act on. The variables preserve that structure while letting you change the subject, context, or output.