If you prompt Muse like ChatGPT, you will get decent answers. If you prompt it like you are delegating to a sharp new assistant, you will get work done. The difference is mostly about what you include: the outcome, the limits, and the moment it has to stop and check with you.
This guide gives you a framework, a template, and real before-and-after examples.
The three kinds of Muse requests
Almost everything you ask Muse falls into one of three buckets. Naming the bucket in your prompt helps it plan.
1. One-off tasks. Do something once and report back.
Find a dentist near 60614 that takes Delta Dental PPO and has a Saturday opening in the next two weeks. Give me the top 3 with times. Don't book anything.
2. Recurring jobs. Do something on a schedule, forever (or until you say stop).
Every weekday at 6:30am, check my calendar and inbox and send me a briefing: today's meetings, anything I need to prepare, and the 3 emails that most need a reply. Keep it under 150 words.
3. Ongoing goals. Own an outcome over weeks, check in, and adjust.
Goal: launch a weekly newsletter by November 15. Build a plan with milestones, check in with me every Monday, and tell me when I'm falling behind.
Recurring jobs and goals are where Muse pulls ahead of chatbots, because it keeps working in the background after you close the app. If you find yourself asking the same question every day, turn it into a recurring job.
The 5-part prompt template
You do not need all five every time, but the more an action matters, the more of them you should include.
OUTCOME: [What "done" looks like] CONTEXT: [Who it's for, why, anything Muse can't guess] CONSTRAINTS: [Budget, dates, sources to use or avoid, tone] STOP POINT: [Exactly when to pause and ask me] FORMAT: [Table / PDF / 5 bullets / decision memo]
Here it is filled in:
OUTCOME: Book dinner for my parents' anniversary. CONTEXT: 4 people, Saturday Nov 7, they love Italian, my dad uses a wheelchair. CONSTRAINTS: Under $80/person, within 20 minutes of Lincoln Park, 6:30 to 7:30pm. STOP POINT: Show me your top 3 options before booking anything. FORMAT: A short table with name, price range, accessibility notes, and available time.
Before and after
| Vague | Better |
|---|---|
| "Help me with my emails" | "Read my inbox from the last 24 hours. List emails that need a reply from me, sorted by urgency, with a one-line suggested reply for each. Don't send anything." |
| "Plan a trip to Austin" | "Plan a 3-day Austin trip Nov 20-23 for 2 adults, $1,500 total excluding flights. Prioritize live music and BBQ. Give me a day-by-day PDF. Ask before booking." |
| "Track my spending" | "Make a spreadsheet of my monthly spending by category from these statements. Flag any category more than 20% above last month." |
| "Review this PR" | "Summarize this pull request as a decision memo: what changes for users, cost, risk, and whether you'd approve. No code diff." |
The pattern: the better versions say what to produce, what to leave alone, and what shape the answer should take.
Always define the stop point
Muse already asks before sensitive actions. But "sensitive" is its judgment, not yours. Stating your stop point removes guesswork:
- "Draft it, don't send it."
- "Fill in the form but stop before submitting."
- "Find options, don't book."
- "Add it to my cart and show me the total."
For anything identity-sensitive, like a login that needs a code from your phone, split the work: you do the one trust step (approve the sign-in), and let Muse handle everything around it.
Ask for decisions, not data
A common mistake is asking Muse to "research" something and getting back ten paragraphs. Instead, ask for the decision you need to make, with a recommendation at the top:
I'm choosing between three standing desks (links below). Give me a decision memo: your recommendation in the first line, then the 3 trade-offs that matter most for someone 6'2" who works from home. Under 200 words.
This works for almost anything: contractor quotes, apartment listings, insurance plans, code changes.
Turn every correction into a standing rule
This is the habit that makes Muse feel smarter every week. When you fix something, do not just fix it once. Tell Muse to make it permanent:
That briefing was too long. From now on, every briefing you send me is 5 bullets max with the most important item first.
You used the wrong price again. Standing rule: always use the price on the checkout page, never the one in search results.
Muse saves these as memories and applies them to future work. After a few weeks, you have a personal rulebook you never had to write in one sitting.
Write your taste down once
If Muse writes anything for you (emails, posts, newsletters), give it your style guide one time and reference it forever:
Save this as my writing style and use it for anything you draft for me: - Short sentences. No jargon. - Lead with the point. - Positive, never snarky. - Bullets over paragraphs when listing 3+ things. - Never use the words "excited" or "thrilled".
Ask for the right format
Muse can produce real files, saved to your Library tab, not just chat text. Ask for the format you will actually use:
- Table for comparisons
- PDF for something you will print or share
- Spreadsheet for anything you will update over time
- Image for a quick visual or draft logo
- Podcast (a short audio briefing) for something you want to listen to on a commute
Use side chats to keep context clean
If you are planning a wedding in one thread and debugging your budget in another, use side chats (Chat tab > menu > pencil-and-note icon). Muse still knows about all of them, but each conversation stays readable.
Quick reference
- Name the type: task, recurring job, or goal.
- Describe the outcome, not the steps.
- Add constraints up front to avoid back-and-forth.
- Always say where to stop.
- Ask for a decision and a format.
- Make corrections permanent.
Want to see these ideas running as full systems? Read 10 Meta Muse workflows that run on autopilot.
Frequently asked questions
Do I need special prompt syntax for Muse?+
No. Write the way you would brief a capable assistant. What matters is stating the outcome, the constraints, where to stop for approval, and the format you want back.
Should I start a new chat for every task?+
Not necessarily. Muse remembers context across chats, and follow-ups work better than starting over. Use side chats to keep separate topics tidy without losing that context.
How do I make Muse remember a preference permanently?+
Say so explicitly: 'From now on, always...' Muse saves standing preferences as memories and applies them to future tasks. You can ask it to list its memories to confirm.
Why does Muse ask so many clarifying questions?+
Because agents act, a wrong guess is more costly than with a chatbot. You can reduce questions by including constraints up front, or tell Muse which details it can decide on its own.