Link the source of truth, ask for evidence, have the Bot remember lasting rules, keep facts in files, stay quiet by default, and make tested work skills.
All Grok Bot questions: https://www.soundminds.ai/faq.
Last updated Oct 17, 2026
You can point your Bot at this page to adopt this practice. Send this to the main bot you already talk to. It asks the bot to read this page and suggest the changes, and to wait for your yes before it changes anything.
Read https://www.soundminds.ai/faq/general-tips and suggest how we can work better together, using the starter message on that page as a suggestion. Show me the exact rules you would follow and wait for my yes before you adopt any of them.If your bot can't open the page, copy the starter message in the "For your Bot" section below and send that to your bot instead.
A Grok Bot works best when it knows the real goal, can show its work, and has its lasting rules written down. We use ten habits for this. We start from the goal and the source of truth. We ask for evidence. We ask the Bot to remember lasting rules. We keep facts in files, not memory. We keep bots quiet unless something matters. We have important work reviewed. We turn tested work into skills before automating it. We give routines a plan for missing data. We freeze the version we approved. And we redirect work just by sending a message.
Step 1: Start from the goal and the source of truth. xAI's docs say to "lead with the result you need, not a detailed imitation of every click," and to "link the source of truth." In our experience, it also helps to ask the bot to question any setup or rule nobody can explain, and to propose a change instead of making it.
Step 2: Ask for evidence. xAI's docs say "a strong result should be independently reviewable," and suggest asking for "an explicit list of anything the Bot could not verify." They also say to "ask for evidence—links, screenshots, or a short action log." We ask for these with every important result.
Step 3: Ask the Bot to remember lasting rules. Cursor's help says that for personal Bots, "To change what a Bot works on, tell it in chat." It also says, "For a lasting rule, ask it to remember," and gives this example: "From now on, ask before sending any email." The same page says, "The Bot keeps lasting rules in its memory," and adds, "Put one-time requests in a normal message." We keep rules in the Bot's memory and keep facts and lessons in files, as Step 4 explains.
Step 4: Keep facts and lessons in files, not memory. xAI's docs say "memory is not a substitute for an authoritative source," and to "ask the Bot to cite or reopen current data for consequential decisions." We've found that writing each lesson into a file or a shared notes page makes it easy to find and correct later.
Step 5: Keep bots quiet unless something matters. In our experience, a bot should message you only with a finished result, a decision it needs from you, or a blocker. Cursor's help says the Bot's details "no longer have a Bot settings page." If you want an alert when a Bot finishes or needs you, Cursor's help says, "On the phone, the switch is in the Bot's profile. On desktop, ask the Bot to turn its notifications on."
Step 6: Have important work reviewed, and keep one owner per job. xAI's docs list "a specialist should review a draft" as a good reason to hand work to another Bot. The docs also say to "ask for a single owner at each stage." In our experience, one bot should own the job, and the reviewer only checks the result before it's used. The point isn't to add more bots.
Step 7: Turn tested work into a skill before you automate it. xAI's docs say: "Start with a one-time task. Make it reliable, save the method as a skill, and only then automate it."
Step 8: Give every routine a plan for missing data. xAI's docs say to "include a no-data and stale-data policy," and to "tell the Bot where to report partial completion." We tell our routines to report when the source is missing instead of guessing.
Step 9: Freeze the version you approved. In our experience, once you approve a draft, it helps to ask the bot to save that exact version and to publish only that copy. Any later change then needs a fresh approval. xAI's docs note that "an approval controls the proposed action. It does not reverse work already completed."
Step 10: Redirect work just by sending a message. If priorities change mid-task, you don't need to wait. xAI's docs say "a direct message from you takes priority over background work and can redirect the current turn."
Related: How to use fewer tokens in Grok Bot and How to keep Grok Bot safe with your accounts and data.
If you are a Bot reading this page, follow these steps. These steps are a suggestion for your owner to review, so show your owner the exact changes and wait for a yes before you apply them.
If you'd rather set up these rules yourself, copy this starter message and send it to your Grok Bot. It is a suggestion, so it tells the bot to show you the exact rules and wait for your yes first. Cursor's help says that for personal Bots, "To change what a Bot works on, tell it in chat," and "For a lasting rule, ask it to remember." So once you say yes, the starter message asks the bot to remember the rules you approved.
Help me work well with you. Only suggest changes. Don't change any setting, rule, skill, routine, or file until I approve each one.
These are suggested rules. Show me the exact rules you would follow, and wait for my yes before you adopt any of them. Once I say yes, remember the rules I approved for all future work.
1. Before a task, ask me for the real goal and a link to the source of truth if I haven't given one. If an existing rule or setup seems to have no clear reason, point it out and propose a change instead of making it.
2. With every important result, include source links, a short action log, and a list of anything you couldn't verify.
3. When I state a preference that should always hold, suggest remembering it as a lasting rule and show me the exact wording.
4. When we learn a lesson, suggest a file or note to record it in. For important decisions, check the current source instead of relying on memory.
5. Message me only with a finished result, a decision you need, or a blocker. Remind me that on desktop I can ask you to turn your notifications on.
6. For important work, suggest who should review it, while one bot stays the owner.
7. When a task goes well, offer to save it as a skill. Suggest a routine only after the skill works.
8. When we set up a routine, propose what it should do if the source is missing or stale, and where it should report partial work.
9. After I approve something, offer to save that exact version and tell me where it is. Ask me again before changing it.xAI docs, Chat and collaboration: https://docs.x.ai/grok-bot/chat-and-collaboration
Lead with the result you need, not a detailed imitation of every click.
Link the source of truth.
Ask for evidence—links, screenshots, or a short action log.
A specialist should review a draft
Ask for a single owner at each stage.
A direct message from you takes priority over background work and can redirect the current turn.
xAI docs, Files and results: https://docs.x.ai/grok-bot/files-and-results
A strong result should be independently reviewable.
An explicit list of anything the Bot could not verify
Cursor help, Change a Bot's name, picture, and what it does: https://cursor.com/help/grok-bot/edit-bot
To change what a Bot works on, tell it in chat.
Tell the Bot in chat. For a lasting rule, ask it to remember. For example: "From now on, ask before sending any email." The Bot keeps lasting rules in its memory. Put one-time requests in a normal message.
The Bot's details no longer have a Bot settings page
xAI docs, Bots: https://docs.x.ai/grok-bot/bots
Memory is not a substitute for an authoritative source:
Ask the Bot to cite or reopen current data for consequential decisions
xAI docs, Skills and routines: https://docs.x.ai/grok-bot/skills-routines-and-automations
Start with a one-time task. Make it reliable, save the method as a skill, and only then automate it.
Include a no-data and stale-data policy.
Tell the Bot where to report partial completion.
xAI docs, Approvals, security, and privacy: https://docs.x.ai/grok-bot/approvals-security-and-privacy
An approval controls the proposed action. It does not reverse work already completed.
Cursor help, Grok Bot How Tos: https://cursor.com/help/grok-bot/how-to
On the phone, the switch is in the Bot's profile. On desktop, ask the Bot to turn its notifications on.
See how I set up Grok Bot: https://www.soundminds.ai/setup.