Independent community · Not affiliated with Meta
Muse Playbooks turns real experiments with Muse into reusable playbooks — for job hunting, freelance work, digital products and local services. No hype decks. Every method comes with steps, a quality check and a kill rule.
Press Ctrl / Cmd + K anywhere to search every playbook, skill and article.
Start here
Muse is an agent that does multi-step work — research, drafting, building and checking — not just a chat box. Learn the five words that run everything here: task, playbook, skill, dashboard, cron.
Muse is in early access, so there is no open signup for everyone yet. Check if your country is live, use the official signup or waitlist for your region with your real details, and if you get in, create your account first, then redeem an invite code in Settings within 48 hours. Full steps are on New to Muse.
How to access Muse, step by step → · Availability Checker → · Access guides →
Pick one small, real job from Tasks, copy the prompt, fill in the brackets, and run it end to end — including the QC check. Then log it in a tracker so the next run builds on it.
Directory
Pick by where you are, not by what's trending.
How to access Muse if you do not have it yet, what Muse is, what a playbook and a skill actually are, and 4 steps to your first useful result.
Step-by-step methods for Job Hunt, KDP Puzzle Books, Adobe Stock and Local Services.
Everything Muse can do, in one filterable grid: Photo Generation, Puzzle Books, Documents, Job Hunt, Stock Media, Local Services, Research, Decide, Plan, Money Ops, Automation, Daily Life & Home, Travel & Places, Shopping & Food, Learn & Translate, Media & Audio, Build & Data. One copy button per prompt.
What Muse uses to do the work: image generation, documents, dashboards, research, crons, memory and connectors, in one filterable grid.
Guides by category: Basics, Tutorial, Workflow, Comparison, Safety, Cost, Access, Settings and News. Each one opens with a direct answer, then steps.
Six reusable skills that save the setup work every time you repeat a job.
Compare hustles in 30 seconds, the money system, and a 7-day start plan.
Dated Daily News, newest first, with source links and a “how this affects a Muse user” line on every item. Plus field notes: what worked, what failed, and why.
The whole start, top to bottom on one page with no dropdowns to open. Access, first task, numbers, tracking, and what to do when you are stuck.
A job Muse does on its own: keep one ledger per business, log every payment and expense the day it happens, hold the tax set-aside, and close the month. A dashboard can be built from the ledger to read it faster — that is one dashboard example, not the only one.
The general guide: what a dashboard is, what a cron is, and how to build them for any lane. Plus a list of dashboard examples — taxes, job hunt, publishing, local work — each one linking to the content that sets it up.
Full-length prompts for experienced users: phased instructions, scoring rubrics, exact output formats and QC gates for job hunt, KDP, stock batches, local services, freelance proposals, research, document production and audits. Plus step-by-step advanced guides, including setting up and applying on any freelance platform.
Fast start
Use the Hustles table. Choose by time to first dollar, not by hype.
Write down three numbers: expected dollars, time to first dollar, and hours per week you can honestly give it. If you cannot put a real number on the first one, it is a hobby — treat it as one.
Playbook = a method with steps, a QC check and a kill rule. Skill = that method saved so Muse reuses it.
A playbook is how a job gets done. A skill is the playbook written as saved instructions Muse follows on every run, so your floors, formats and checks survive between sessions.
Pick the playbook for your lane and do step 1 today. Small and shipped beats planned and perfect.
Complete the first step of a playbook before you read anything else on this site. Then make a share card with the Share Card Generator in Tools and share it anywhere you already are.
Changelog
News now tracks Muse itself.
Daily News · Last updated Nov 3, 2026, 6:00 AM ET
Newest first on the News page: Muse news first, broader AI that affects Muse users second — every item dated, sourced, and with a plain “how this affects you” line.
Referral code
Just joined Muse, or about to? Copy this code and redeem it in the app to claim the invite tokens.
Small print: GZTOI6 is a Muse invite/token code, not cash. A new user usually has to redeem it within 48 hours of joining, in Settings > Redeem token in the Muse app.
Featured
Remote, hybrid or onsite. Set a pay floor, verify on the employer's own page, QC before you submit. Never guess a screening answer.
Test demand before you build. Solver-checked puzzles, thumbnail-first covers, and a strict niche test.
FAQ
Short answer: Muse is in early access and rolls out by country, so you cannot just download it and sign up everywhere yet. Do this, in order:
An invite code on its own does not create an account. You join Muse first, then you redeem the code inside your new account.
Meta's personal AI agent. You give it a job and the context it needs, it works through the steps — research, draft, build, check — and brings back a result for you to approve. It is built for jobs with steps, not just quick questions.
No. Muse Playbooks is an independent community resource, not affiliated with or endorsed by Meta. For official announcements, use Meta's own pages — for example the Connect 2026 announcement.
No. Most Muse Playbooks lanes — job hunt, documents, local services, publishing — need a clear job, honest numbers and a QC habit, not code. If a task does need a technical step, Muse explains it in plain language and you approve it before it runs.
If you do not have Muse yet, start with the access answer above: check your country, use the official signup or waitlist, and redeem a code only after your account exists. If you already have Muse, do the 3 steps at the top of this page: understand what Muse is, confirm your plan and limits in Settings, then do one small real task from Tasks and log it in a tracker. One finished task teaches more than an hour of feature reading.
Check your own account. Pricing, token allowances and limits change during early access and differ by plan and region, so Settings in your Muse account is the only reliable source. The Token Estimator in Tools helps you size a job before you run it.
No. An invite code pays in Muse tokens — credit to use Muse — not cash. Redeem it promptly after joining, in Settings > Redeem token; the usual window is 48 hours. See the Redeem Code guide.
It should not, the way Muse Playbooks sets it up. Drafts, proposals and applications are staged for your approval, and approval is per action — approving one message does not approve the next. Crons log, draft and alert; they do not send, pay or submit.
Rollout is staged by plan and region. Announced is not live: a feature counts for your workflows only when it appears in your own account and you have run it once. Until then, keep the manual way of doing that job.
One tracker per lane, one pinned dashboard built from it, and a cron that reconciles it on a schedule. Chat history is not a tracker. The full setup is on Crons & Dashboards. For money specifically, the ledger itself is in Money Ops.
For beginners
You don't need to be a developer. You need a clear job, good context, a habit of checking the output — and a place to track it, or you will lose the thread by week two.
First question
If you do not have Muse yet, start here. Muse is in early access. That means access is not open to everyone, everywhere, at the same time. It rolls out by country and by account, so the steps below are the order that works.
Check if Muse is live in your country, go through the official signup or waitlist for your region with your real details, create your account when you are let in, then redeem an invite code inside the app if you have one. If you want to use mine, it is GZTOI6. Note: GZTOI6 is a Muse invite/token code, not cash, and a new user usually has to redeem it within 48 hours of joining, in Settings > Redeem token in the Muse app. There is no separate Muse Playbooks signup, and an invite code by itself does not give you Muse.
No signup option for your country? That is the Not yet case: waitlist once, real details, re-check after announcements.
Code will not redeem? Check spelling, check you are still inside the 48 hour window, and check the code has not already been used on this account. Then use official support. Do not buy another code from a stranger.
Signed up but a feature is missing? Check your own account again after announcements. Announced is not live.
Someone offers to sell you access? Do not buy it. Access tied to someone else's account or region can disappear, and you lose anything built on it.
An email address you can check for confirmations and waitlist news. Details that match the country you actually live in. A few minutes to finish Settings before your first real job. Nothing else on this page works until this part is done, so do not skip ahead to playbooks if you cannot sign in yet.
Walkthrough
Do this once, in order, the first time you sign in. One finished, checked result teaches more than an hour of feature tours.
Troubleshooting
Most new-user blocks are one of these eight. Find your case, do the fix in order, and only move on when the check passes.
Likely cause: your country is Not yet, or you are on a waitlist. Fix: check the Availability Checker, join the official waitlist once with real details, and re-check after announcements. Do not use a VPN or borrowed address. Check: you can state clearly whether you are Live, waitlisted, or Not yet.
Likely cause: typo, past the 48-hour window, or already redeemed on this account. Fix: retype GZTOI6 exactly in Settings > Redeem token, check the window, check your balance history, then use official support. Never buy a replacement code from a stranger.
Fix: ask for sources, open the source yourself, and re-run with “verify against this source and flag guesses.” For prices, dates and job details, the source wins every time.
Likely cause: no context, no numbers, no format. Fix: add your floor or budget, your source material, an example of good, and a done definition. Vague in, vague out is not a Muse failure.
Likely cause: announced, not live for your plan or region yet. Fix: check your own account, keep the manual way of doing that job, and re-check after announcements. Do not rebuild a workflow around it.
Fix: stop at the first confirmed block. Confirm once in a normal browser, log it once with the date, and reroute: manual use, official alerts or API, or another platform. No bypass tools, solvers, or retry loops.
Likely cause: work was done in chat and never logged, or a cron stopped. Fix: reconcile today from mail and notes, stamp the new reconciled date, and set one weekly sweep cron that reports only gaps.
Fix: size the job first in the Token Estimator, cut the prompt to one job per run, cap reply length, and check your balance in Settings before any multi-step run. Small finished jobs beat one giant run.
Muse is Meta's personal AI agent. Beyond answering questions, it can research, draft, build pages and documents, work through multi-step tasks, and use skills — reusable instructions for jobs you repeat. You stay in charge: you approve sends, payments and submissions.
A playbook on Muse Playbooks is not a vague guide. It has: the goal, the setup numbers, numbered steps, a QC check before you ship, and a kill rule that tells you when to quit a lane that isn't paying.
A skill is a playbook Muse itself follows. Instead of re-explaining your rules every time — your price floor, your QC list, your format — you save them once as a skill and Muse applies them on every run.
A dashboard is built from a tracker or ledger — applications, income, books, batches, jobs — and pinned so it is the first thing you open. It keeps no numbers of its own: every total traces to the source underneath it, and it shows when it was last updated.
A cron runs whether or not you open a chat: a weekly sweep, a deadline watch, a payment alert. It updates your tracker or dashboard and reports only what needs a decision. Crons draft, log and alert — they never send, pay or submit without your go.
Connectors link accounts Muse can work with, like email, calendar and files — only connect what a real job needs, and test that the connection is live. Memory is standing context saved between sessions: your floors, formats and rules, so you stop re-typing them.
A good answer once is not a system. Memory and skills are what make it repeat.
New users re-explain the same floor, format or preference in every chat and wonder why results drift. The fix is boring: say the rule once, ask Muse to confirm exactly what it saved in one line, and put repeat jobs into a skill. If a rule matters — a pay floor, a price, a format, a no-guessing rule — and it is not saved, assume the next run does not have it.
Skills Library → · Guide: memory, connectors and permissions →
Check the source, and check the finished file — not the chat reply.
Prices, dates, job details and program terms go stale or get copied wrong. For anything you will act on, verify at the source: the employer's own page, the official listing, the actual payout record. And when Muse builds a document or file, open the exported file and read it — generated files can quietly truncate text or misalign numbers while reporting the right page count. Every playbook on this site has a QC step for this reason. Do not skip it to save five minutes.
Drafts stage and wait. Nothing should send, pay or submit on autopilot.
Set Muse up so replies, proposals, applications and purchases are drafted and staged for your approval, not sent. Approving a draft once does not approve the next one. Never hand over passwords, payment card numbers, or one-time codes for accounts you are not actively signing into at Muse's request, and keep other people's private details and client data out of anything public.
A feature shown at an event is not a feature in your account until you can run it.
Muse is in early access. Capabilities roll out by plan and region, and announcements describe direction. Muse Playbooks labels features Live, Announced or Reported for this reason. The test is always your own account: if the setting or channel is not there, it has not reached you yet — build on what is live and keep a manual fallback. That applies to Connect 2026 announcements too.
Guide: Connect 2026 and Muse Updates → · Muse Updates in News →
Bot walls like Cloudflare are a stop sign for that route, not a puzzle to beat.
Freelance and job platforms sit behind bot management. Automated access can trip a "Verify you are human" wall, looping CAPTCHAs, or silent failures. Stop at the first confirmed block. Do not use bypass tools, CAPTCHA solvers or retry loops — that can get the account flagged or banned. Park that route, log it once with the date, and work the same lane manually, through official alerts or an API, or on another platform. Keep your portfolio and tracker in your own files so a parked route strands nothing.
And invite tokens are working credit, not income.
Plan pricing, token allowances and reset timing change during early access and differ by region. Before you budget a workflow around a limit, check Settings in your own Muse account. An invite or referral code pays in tokens to use inside Muse — useful for experiments, never count it as cash in a hustle plan.
Chat history is not a tracker. This is the mistake that kills most lanes quietly.
New users run good work and keep it in scattered chats: applications with no status, income with no ledger, batches with no results. Two weeks later nobody can say what was sent, what paid, or what is overdue. Set up tracking on day one, before the work piles up: one tracker or ledger per lane, one pinned dashboard that reads it, and a cron that keeps it current. That system is covered in full on its own page.
Crons & Dashboards: build the tracking system → · Money Ops for income and taxes →
Tick each day off as you finish it. The progress bar is for this session only — if you want it to survive, copy your finished list into your tracker.
Full setup: Crons & Dashboards → · Prefer no panels? Read Quick Start straight through →
Task: copy a Job Hunt prompt from Tasks. Playbook: run Job Hunt steps in order, QC before submit. Tracker: every application logged with status and next step. Dashboard: pinned view of live applications, interviews and deadlines. Cron: weekly check that flags follow-ups due and closing dates — you still approve every send and submit.
Skill: your lane's rules saved once. Ledger: every payment and expense logged the day it lands, set-aside moved. Dashboard: pinned Money Ops view — income by lane, set-aside owed vs held, deadlines. Cron: weekly expense sweep reconciles it and flags only gaps. Close: monthly numbers you can trust, because they were kept current all month.
Copy one, replace the brackets with your situation, and run it.
Experienced user? The starter prompts above are the short versions. Open the advanced prompts library →
Directory
Four tested methods. Each one is its own page with full steps, a QC check, a kill rule, a deep dive with templates and numbers, and a Verified in the wild section covering what other Muse users have done — labeled tested, reported, or announced.
Remote, hybrid or onsite. Pay floor first, total-pay comparison, QC before every submit.
Validate the niche with sales-rank tests, build solver-checked books, design for the thumbnail.
Small weekly batches, 100% zoom QC, and a clean CSV pipeline for titles and keywords.
Proof first, quote before work, deposit up front, and one review ask after a finished job.
Playbook · Career
Works for remote, hybrid and onsite roles. The method is the same: define the job you can accept before you look, then protect yourself with QC.
Log every application in the tracker above, pin a dashboard of live applications, interviews and deadlines, and set a weekly cron to flag follow-ups due and closing dates. Drafts still wait for your approval. Crons & Dashboards →
Application tracker fields, total-pay worksheet, and the screening-answer rule.
Company · role · source · date applied · pay range · work type · status · next step · date of next step. Before applying anywhere, search this tracker and your email for the company — double applications look careless.
Total-pay worksheetBase pay + employer retirement match (check vesting and plan-year rules) − health premium difference − surcharges + PTO value + schedule value. Compare a full year on both sides, not the offer letter’s first line.
Screening answersWrite your standing answers down once (authorization, sponsorship, location, clearance, relationships) and reuse them verbatim. If a question is not covered, stop and find out — a guessed answer can end a candidacy or worse. Also agree your reason for leaving and earliest start date with yourself in advance, so every application tells the same true story.
What other Muse users — not Muse Playbooks — have actually done, labeled by how solid the evidence is.
How to read the labels: Tested — an independent user ran it and published what happened. Reported — the user says they use Muse this way; nobody independent has checked the results. Announced — from Meta's own announcement, not a user test.
Tested Meeting preparation, done before you walk in. An independent tester who ran 10 real business tasks through Muse used it to prepare for meetings — pulling the relevant context together ahead of time instead of digging through threads in the room. The same move works for interviews: brief yourself on the company, the role and your own talking points before you sit down. Source: Independent tester — 10 business tasks.
Tested A competitor brief that finished after the app was closed. In the same test, a competitor brief was still completed after the tester closed the app — the job ran to the end on its own and the finished brief was waiting. For a job hunt, that is employer research you can queue and collect later: set it running, go apply, come back to a finished brief. Source: Independent tester — 10 business tasks.
Reported Weekly performance briefings instead of daily checking. Creators report having Muse pull their numbers together into one weekly briefing rather than opening dashboards every day. A job hunt runs the same way: one weekly rollup of applications sent, replies in and follow-ups due beats re-reading the tracker daily. Source: Creator social media and brand deal tracking.
Playbook · Publishing
Puzzle books look passive. The winners are the ones who validated demand and checked every page.
Keep a pipeline tracker — themes tested, verdicts, books in production, QC results — pin a dashboard of what is live and selling, and let a weekly cron surface stalled books and untested themes. Crons & Dashboards →
How to read a sales rank, and the cover spec that survives the thumbnail.
Lower is better: a rank in the low tens of thousands means steady daily sales; hundreds of thousands means occasional sales; over a million usually means effectively none. A strict pass wants most of the top 5 competing titles under ~100,000, with recent publication dates and real review counts. One outlier hit among dead titles is luck, not a niche. Third-party “sales estimate” tools are model guesses — label them that way.
Cover specTitle set in a bold sans-serif, readable at 200px wide; one clear theme image (no clip-art clutter); high contrast; puzzle count and “Large Print” badges in corners a thumbnail keeps. Test at thumbnail size before upload — that is the size buyers actually judge.
Interior specPuzzles generated and solver-verified 100%; solutions included; instruction page read in full in the exported file. If cover art is AI-generated, disclose exactly that element in KDP.
What other Muse users — not Muse Playbooks — have actually done, labeled by how solid the evidence is.
How to read the labels: Tested — an independent user ran it and published what happened. Reported — the user says they use Muse this way; nobody independent has checked the results. Announced — from Meta's own announcement, not a user test.
Tested A Facebook caption written from the page's own track record. An independent tester running 10 real business tasks through Muse had it write a Facebook caption based on how the page's earlier posts had actually performed — not a generic caption written blind. For a puzzle book launch, that means promo posts built on what your own audience already responded to. Source: Independent tester — 10 business tasks.
Reported Caption and Reel ideas mined from Instagram and Threads. Creators report having Muse analyze their Instagram and Threads activity and turn it into caption and short-video ideas. For KDP, that is a repeatable way to get book promo angles out of your own accounts instead of inventing posts from nothing. Source: Creator Instagram and Threads analysis test.
Reported A weekly briefing on what performed. Creators also report getting one weekly performance briefing from Muse — what landed, what did not — instead of checking stats daily. Applied to publishing: one weekly look at which books and posts moved, then decide the next book or post from that. Source: Creator social media and brand deal tracking.
Playbook · Creative
A background compounder, not a lottery ticket. Small, consistent, quality-checked batches beat one giant dump.
Log every batch in a tracker — assets, QC scores, approvals, downloads by category — pin the results on a dashboard, and run the weekly batch on a cron reminder so the rhythm survives busy weeks. Crons & Dashboards →
The exact shape of a batch that uploads cleanly the first time.
Rows are subjects, columns are settings, layers are styles. Fill cells deliberately so the batch covers demand without near-duplicates competing with each other in search.
CSV columnsfilename · title · keywords · category — one row per asset, built during QC, not reconstructed afterwards. Titles describe what a buyer would search for; keywords stay relevant and honest.
RhythmSame day weekly, batch capped around 20. Track approvals, rejections and first downloads per category; two flat batches in a category is data — rotate.
What other Muse users — not Muse Playbooks — have actually done, labeled by how solid the evidence is.
How to read the labels: Tested — an independent user ran it and published what happened. Reported — the user says they use Muse this way; nobody independent has checked the results. Announced — from Meta's own announcement, not a user test.
Reported Caption and Reel ideas from Instagram and Threads analysis. Creators report having Muse work through their Instagram and Threads activity and come back with caption and short-video ideas based on it. For a stock contributor, that is the marketing side of the lane: regular posts about new batches, generated from what your own accounts show is working. Source: Creator Instagram and Threads analysis test.
Tested A Facebook caption built on actual page performance. An independent tester who ran 10 real business tasks through Muse had it draft a Facebook caption using the page's real past performance as the input. Promoting a new batch the same way — from your own numbers, not from a template — is the version worth copying. Source: Independent tester — 10 business tasks.
Reported One weekly performance briefing. Creators report letting Muse assemble a weekly briefing on how their content performed, instead of checking dashboards daily. The stock parallel is direct: one weekly rollup of approvals, rejections and downloads by category is exactly what the batch tracker in this playbook is for. Source: Creator social media and brand deal tracking.
Playbook · Local
The fastest lane to first dollars for most people: a useful skill, sold to neighbors and local businesses, with proof doing the marketing.
Keep a jobs tracker — inquiry, quote, deposit, booked, done, reviewed — pin a dashboard of quotes waiting and jobs this week, and use payment alerts plus a weekly sweep cron so no finished job misses its one review ask. Income from every job goes in the ledger. Crons & Dashboards → · Money Ops →
The written bits that stop local work going wrong.
Job · flat price · deposit · what is included · what is not · parts policy (who buys, any sourcing fee agreed in advance) · timing. If it is not in the quote, it is a future argument.
Scope honestyKeep a public in-scope / out-of-scope list and honor it in replies. Referring out work you cannot do safely or legally protects the reviews that feed everything else.
Review systemOne ask, after a finished happy job, with both review links in a single message. Log who has reviewed you and never ask them again. Pitch partners only where the need is visible on their own public presence.
What other Muse users — not Muse Playbooks — have actually done, labeled by how solid the evidence is.
How to read the labels: Tested — an independent user ran it and published what happened. Reported — the user says they use Muse this way; nobody independent has checked the results. Announced — from Meta's own announcement, not a user test. The four pilots below are Meta's own examples, told by Meta — treat the outcomes as the company's account, not an independent audit.
Announced A grooming company read 87,000 reviews in minutes. Meta's pilot account: a pet grooming company had Muse analyze roughly 87,000 customer reviews and produce a strengths-and-weaknesses analysis in minutes — work that would take a person weeks. For a local service, that is your own review profile, mined: what customers praise, what they complain about, in one sitting. Source: Small business pilots — Bennett Orchards and Live Bearded.
Announced A farm operator using Muse as the office help. In Meta's pilots, a farm operator uses Muse as a general assistant for the tasks around the real work, freeing their own hours for the work only they can do. That is the local-services pattern exactly: the owner does the job, Muse carries the admin around it. Source: Small business pilots — Bennett Orchards and Live Bearded.
Announced A grocery owner putting Muse on back-office roles. Meta describes a grocery store owner using Muse across back-office jobs — the paperwork, ordering and admin layer of a small shop. If you run a small local operation, that layer is where your evenings go; it is also the layer most safe to hand over first. Source: Small business pilots — grocery and boutique.
Announced A boutique owner who had Muse audit the email automations. In Meta's account, a boutique owner had Muse go through their existing email automations, audit what was actually running, and rebuild the flows with tracking added. For a local business with a neglected email list or follow-up flow, an audit-and-rebuild is a concrete first job to hand over. Source: Small business pilots — grocery and boutique.
Reported Deal and inquiry email, tracked instead of buried. Creators report having Muse track their brand-deal email — what came in, what is waiting, what is owed. Translate that to local work and it is quote requests and follow-ups: the inquiries that decide whether a week is booked rarely get a system, and they should. Source: Creator social media and brand deal tracking.
For experienced users
Full-length prompts with the context, phases, scoring, output format and QC gates built in. Replace every [bracket] with your real inputs before you run one. Step-by-step guides are below the prompts.
Score, verify, tailor and QC multiple applications in one run, under a no-guessing rule.
Advanced
Replace first: your resume, pay floor, work type, and standing screening answers.
Scores the job post first, turns ambiguities into questions, then writes a proposal a busy client can approve in one read.
Advanced
Replace first: the full job post, your 2–3 closest proof samples, your price, and your scope cap in hours.
The whole book pipeline in one prompt, with a hard validation gate before any building starts.
Advanced
Replace first: the theme, the audience, and the format (for example, large-print word search).
Matrix planning, a scored 100% zoom QC rubric, and the upload CSV, run as one batch process.
Advanced
Replace first: subjects, settings, styles, and the batch cap (20 is the standard ceiling).
Scope check, written quote, booking messages and the one-time review ask, all under your service rules.
Advanced
Replace first: services offered, out-of-scope list, service area, deposit, and both review links.
A full document build with a fixed structure, strict sourcing rules, and a QC pass reported line by line.
Advanced
Replace first: the task, audience, source material, system and version, and the format you need.
Evidence with sources and dates, bull case vs bear case, unit economics, and a kill metric set before you start.
Advanced
Replace first: the idea, niche or purchase you are deciding on.
Converts a repeated job into saved instructions with triggers, approval gates, QC and failure modes — then test-runs it.
Advanced
Replace first: a description of the job, or notes from the last 3 times you did it.
An auditor prompt that hunts invented claims, mismatched numbers and silent truncation, and fails loudly.
Advanced
Replace first: the document, application, file or page to audit, and the source it was built from.
Reconciles the ledger against real deposits, recomputes the set-aside line by line, and closes the month with a defect table.
Advanced
Replace first: the month, your ledger, payout or deposit records, set-aside balance, and set-aside percentage.
Workflows, not prompts. Written generally, so each one works on whichever platform you choose.
The full setup on any freelance marketplace: account, profile, portfolio, then applying the right way.
Guide
One master portfolio you can drop into any platform profile, proposal or page without rebuilding it.
Guide
Task library
Pick a filter, copy a prompt, replace the brackets with your details.
Turn this selfie into a clean professional headshot. Keep my face and features exactly as they are, replace the background with a plain studio backdrop, fix the lighting, and give me 3 crops: square, portrait, and wide.
Place this product in a [kitchen / desk / garden] scene a buyer would picture using it in. Natural light, no text, no extra logos, and leave empty space on the [left / right] for a headline.
Turn this photo of my dog into a painted-style portrait. Keep their markings and expression exactly, use a [dark / light] background, and deliver it at print size with room to crop square.
Restore this photo: fix [scratches / fading / tears / blur] without changing faces, clothing or the setting, keep it true to the original, and give me the restored version plus a side-by-side comparison at full size.
Before I build this puzzle book, test the demand: find the top competing titles for [theme], their prices, review counts and sales ranks, then give me a strict pass or fail with the numbers.
Build a [theme] word search book: [90] puzzles at [difficulty], every puzzle solver-checked with a verified solution, a checked word list per page, and solutions at the back.
Design a cover for [title]: bold title readable at 200px wide, one clear [theme] image, high contrast, and badges for [puzzle count] and Large Print where a thumbnail keeps them visible. Show me the thumbnail test.
Turn these messy notes into an SOP a new hire can follow alone: purpose, prerequisites, one action per numbered step, the 3 most likely errors and fixes, and an escalation path. Flag anything missing instead of inventing it.
Turn this meeting transcript into decisions, action items with owners and dates, and a short recap email. Use only what was actually said; list anything unclear instead of guessing.
QC this exported document against its source before I ship it: check every number, name and date, read the instructions page in full, and report defects by page. Pass or fail, no “looks good.”
Rank these job postings against my resume and pay floor of [$X]. Score skills match, pay, and work-type fit, drop anything below the floor, and flag any screening question I would have to guess.
Tailor my application for [role] at [company] using only facts from my resume, reframed in the posting’s language. Then QC it: names, dates, salary, and the exact title and company. Draft only, do not submit.
Prep me for an interview for [role]: likely questions based on the posting, answers built from my real work, and 5 questions I should ask them. Nothing invented.
Plan a stock batch for [subject]: a matrix of subjects, settings and styles with no near-duplicates, capped at [20] assets, plus the title and keyword CSV format I will upload with.
QC these stock assets at 100% zoom: hands, text, edges and artifacts. Score each out of 10, reject anything under 8, and build the upload CSV with filename, title, keywords and category per file.
Turn this customer inquiry into a written quote: whether it is in my scope, the flat price, deposit, what is included and what is not, and the parts policy. Draft the reply in short, plain language. Do not send it.
Write my one review ask for a finished, happy job: a short thank-you with both my Google review link and community recommendation link in the same message. [paste both links]
Before I build [product], test the demand: who is already selling it, at what price, and how well does it rank? Show your sources, then give me a pass or fail verdict with the numbers.
Write a decision memo on [idea]: who already sells it, at what price, the bull case and the bear case with numbers, what has to be true for it to beat my current option, and a GO, TEST or NO-GO verdict, with sources and dates.
Verify this claim: [claim]. Find the original source, check it directly, find one independent source that agrees or disagrees, and tell me confirmed, wrong, or unverified, with links and dates.
Log this payment in my side-income ledger: [source, gross, fees, date]. Move the 30% set-aside on profit, categorize these expenses [list], and show me set-aside owed vs held.
Close my month: reconcile the ledger against payout records, flag uncategorized items instead of guessing, and update my pinned dashboard so every total traces to the ledger.
Check if I can afford [purchase] at [$ price / $ per month]: use my actual income, recurring bills and current balances, show the monthly impact for 12 months, what it delays, and a buy, wait, or skip verdict with the numbers.
Design a safe workflow for [task]. Define the trigger, inputs, steps, checks, failure handling, approval points, and useful output. Nothing sends, pays or submits without my approval.
Turn this workflow I repeat into a reusable skill: trigger, inputs, my rules written as testable lines, steps, output format, a QC checklist, failure modes, and what must wait for my approval.
Run my admin reset: sort this inbox and task list into do now, schedule, delegate and drop. Draft replies for the do-now items, stage them, and give me one list of what needs my decision.
Turn [idea] into a build brief. Define the user, core job, must-have flows, content, edge cases, launch version, and what to leave out.
Search my email for [topic / person / date range]. Summarize each thread in one line, sort it into answer now, waiting on someone else, and no action, draft short plain replies for the answer-now ones, and stage them for my approval. Do not send anything.
Find a [30 / 60] minute slot for [people] between [dates] that works across our time zones, using my calendar's free and busy times. Draft the invite with the agenda and what each person should bring. Stage it, do not send it yet.
Check my [camera / device name] for events since [time]: summarize what triggered, when, and anything unusual, and check whether [light / lock / thermostat] is in the state I asked for. Report only what the devices actually show.
Plan a trip to [destination] for [dates], [number] people, budget [$X]: compare transport, places to stay by total price, and a day plan that fits [interests]. Keep ideas, shortlisted options, and anything booked in separate lists, with a source link for every price.
Compare options for [route / destination] on [dates]: for flights use the delivered total with bags and fees and the total travel time, for stays use the full-stay total with taxes. Rank them, flag tight connections and bad cancellation terms, and link each option.
Find [restaurants / cafes / shops / parks] near [place] that are open at [day and time], rated [minimum] or better. Give me the address, hours and a map link for each, verified from the place listing, not a summary.
Research [product] for [use]: compare the real options by delivered price, key differences, and return terms, with a link to each listing you priced. End with buy, wait, or skip, and the reason in numbers.
Evaluate this listing for [item] at [$ price]: find comparable listings to set a fair price range, list the questions to ask the seller, the red flags in the photos and wording, and the safe way to meet and pay.
Plan [number] days of [meals] for [household size] around [likes, dislikes, budget]: simple recipes that reuse ingredients, then one grocery list grouped by store section with quantities. Flag anything I said I already have.
Teach me [topic] for [goal]: ask me 5 diagnostic questions first, build a session plan from my answers, teach one session at a time with worked examples, quiz me at the end of each session, and keep a list of what I got wrong.
Translate this into [language] for [audience]: give me a natural version and a formal version, keep the tone [casual / professional], and explain any phrase that has no direct equivalent instead of translating it word for word.
Explain this [letter / contract section / report] in plain language, section by section: what it says, what it asks me to do, every deadline and amount, and the parts I should get a qualified professional to confirm before acting.
Make a [30 / 60] second faceless video about [topic] for [platform]: write the script, plan the scenes, generate the visuals, add the narration and captions, and show me the script before you render anything.
Turn my notes on [topic] into a [length] podcast episode: a script with an opening, [number] segments and a close, sources for any claim that needs one, recorded narration, and a title and description written from the finished episode.
Read this text aloud as a finished audio file: pace [calm / brisk], pronounce [names or terms] the way I specify, pause at section breaks, and deliver the audio with a transcript that matches it word for word.
Build a playlist for [activity / mood] lasting about [length]: find the exact recordings of [artists / songs], mix in similar tracks I would not know, order it so the energy builds, and create it in my music app if it is connected.
Build a page for [purpose] from this brief: who it is for, the sections in order, the exact text and images to use, and what it must do on a phone. Build it, check it at phone and desktop widths, and report what you verified.
Clean this spreadsheet of [data]: fix inconsistent names, dates and categories, flag rows you cannot classify instead of guessing, then give me the totals and the top 5 patterns, with the cleaned file as the deliverable.
Debug this [error / broken thing]: reproduce it first, state the most likely causes in order, test the cheapest one first, fix it, and prove the fix by running the same steps that failed before.
Organize these [notes / documents] into the structure I already use: group them by [topic / project], name each file consistently, write a one-line summary per item, and give me an index that links everything. Do not delete anything.
Compare these two offers on total pay, not headline salary: base, employer retirement match and its rules, health costs and surcharges, PTO, schedule and commute. Show a full-year table for both, name what each number is based on, and end with a pick and the one fact that decided it. [paste both offers]
Decide keep, fix or kill for [lane] from these 7-day numbers: [hours spent, shipped, dollars in, dollars out]. Compare them against the kill rule I set at the start, show the math, and give one verdict. If fix, name the single change and what result would kill it next week.
Score these options for [decision] against my criteria and weights: [criterion = weight, ...]. Score each option 1 to 5 per criterion with one line of evidence, total the weighted scores, flag any score based on a guess, and tell me which single fact, if wrong, would flip the result. [list options]
Plan my week for [lane] in the [5 to 10] hours I actually have: [available blocks]. One outcome for the week, tasks in priority order fitted to those blocks, anything that does not fit goes on a not-this-week list. End each day with the one thing that must ship that day.
Turn this goal into a dated plan: [goal, deadline]. Work back from the deadline into weekly checkpoints with a measurable result for each, the tracker fields I will log them in, and the dashboard panels that will show progress. Flag any week that needs more hours than I have: [hours per week].
Before I start [lane / project], set up how I will track it: the tracker fields, the pinned dashboard panels, and one weekly cron that reconciles both and reports only exceptions — overdue steps, missing items, upcoming deadlines. Show me the empty tracker and one example row.
Create a weekly cron for [lane] on [day and time]: reconcile my tracker against [mail / notes / payments], update my pinned dashboard, stamp its reconciled date, and report only exceptions — overdue next steps, missing items, deadlines in the next 7 days. Flag anything ambiguous for my call, never guess, and never send or submit anything.
Full-length versions of these prompts, with phases, scoring and QC gates: Advanced Prompts →
Tool directory
Pick a filter, copy the starter prompt, replace the brackets with your details.
Pick a country to see the Muse Playbooks snapshot for Muse access there, as of November 2026.
Access rolls out in stages and changes without much notice. Your own Muse account and the official signup for your region are the final word — if they disagree with this list, they are right.
What to do next: If your country is Live, go to the official Muse signup or official app listing for your region, create your account, open Settings to confirm your plan, then redeem an invite code in Settings > Redeem token within 48 hours of joining. If it is Not yet, join the official waitlist once with your real details and re-check after announcements. Do not use a VPN or a borrowed address to fake a region.
Rough-size a job before you run it. Paste the prompt you plan to send, then set how big a reply you expect and how often you will run it.
Estimate only: roughly 4 characters or about 0.75 words per token for English text. Research, tools, files and images use tokens differently — check the balance in your own account before a large run.
Five questions. Answer honestly — the result tells you the one thing to fix first.
How long your money lasts at your current pace — and what changes it fastest. Use your real numbers from your ledger.
Runway = cash ÷ (monthly spending − monthly income). If income covers spending, runway is open-ended at this pace — the job then is growing the gap, not cutting. This is a planning estimate, not financial advice.
Pick a goal, add your topic, and get a complete prompt in Muse Playbooks's own structure — context, steps, format, and a QC check built in.
Compare two ways of doing the same job — for example doing it yourself vs using Muse, or two plans — by true monthly cost including your time.
True cost = cash cost + (your hours × your hourly value). The cheaper-looking option often loses once your time is priced in. Check any Muse plan price in your own account — pricing there is the source, as the coherence note on News states.
Generate or edit images: headshots, product shots, pet portraits, book covers and stock batches. Give Muse a reference photo when identity has to stay exact.
Build finished documents, spreadsheets, presentations, PDFs and CSVs to a spec, with real formatting, then QC the exported file itself.
Trim, convert, resize and extract audio from video and audio files. Muse inspects the file first and uses the smallest conversion that works.
Turn finished work into a shareable page: a guide, portfolio, services page or comparison that other people can open from a link.
A pinned page built from your underlying data: income by lane, set-aside owed vs held, open items, deadlines and lane status.
Save a repeated job as reusable instructions: trigger, inputs, your rules, steps, output format, QC checklist and failure modes.
Search a real question, who sells it, at what price and how it ranks, and get the sources back so you can check the numbers that matter.
Break a bigger question into sub-questions, search across sources, cross-check claims against each other, and report what is still unverified.
Open the original page behind a listing, summary or claim and check it directly: title, price, terms and dates at the source.
Jobs Muse runs on a timer: a weekly expense sweep, a set-aside check, a deadline watch. Each one reports only what needs a decision.
Hand Muse a whole job in one run: research, draft, QC and report, instead of copy-pasting between separate chats.
Design a safe workflow for a repeated task before automating it: trigger, inputs, steps, checks, failure handling and approval points.
Log every payment and expense the day it happens, move the tax set-aside on profit, and close the month against real payout records.
One row per item, with status and next step: applications, freelance jobs, stock batches, publishing projects.
Standing context Muse keeps between sessions: your floors, formats, rules and preferences, so you stop re-explaining them every chat.
Connect accounts Muse can then read and work with, such as email, calendar and files. Check the connection is live before relying on it.
Search mail, read free and busy times, and draft replies, follow-ups and invitations that wait for your approval before anything sends.
Open and read public pages in a real browser session when a page needs interaction or a listing has to be checked where it actually lives.
Prompts that use these tools, organized by job: Tasks → · Full workflows: Guides →
Guide directory
Guides by category. Every guide opens with a direct answer, then the steps.
Meta’s personal AI agent, built to do multi-step work, not only answer questions.
Direct answer: Muse is a personal AI agent from Meta. You give it a job and context, it works through the steps, uses tools and skills where they fit, and brings back a result for you to check and approve.
Task is the job, playbook is the method, skill is the method saved for reuse.
Direct answer: A task is one job with a prompt. A playbook is a tested method for a repeated job, with steps, a QC check and a kill rule. A skill is that playbook saved as instructions Muse applies on every run.
One small real job, end to end, beats an hour of reading about features.
Direct answer: In 15 minutes you can get one finished, checked result: pick a small job you already need done, give Muse your numbers and source material, run it, then QC the output before you use it.
Pick one lane, set the numbers, do step 1 today, log keep / fix / kill after 7 days.
Direct answer: A playbook only works in order: numbers first, steps second, QC before you ship, and a dated keep / fix / kill call at the end of the first week.
Four lanes, one pattern: validate, build, QC, ship, log.
Direct answer: Every Muse Playbooks playbook runs the same loop, only the lane changes: set the numbers, validate before building, produce in small batches, QC against the source, ship, and log the result.
The ledger is the system. A dashboard and crons are add-ons that make it easier to read and keep current.
Direct answer: Money Ops starts with one thing: a ledger where every payment and expense is logged the day it happens, with the tax set-aside moved on each payment. That ledger stands on its own. Two add-ons build on it: a dashboard built from the ledger, so you can read the totals at a glance — one dashboard example among several — and crons that reconcile the ledger on a schedule and report only exceptions.
Open Money Ops → · Crons & Dashboards: the general guide and all dashboard examples →
One tracker, one pinned dashboard, and scheduled jobs that reconcile both — for any lane.
Direct answer: Start with a tracker (one row per application, job, book, batch or payment), build one pinned dashboard from it with a last-reconciled date at the top, then set a weekly cron to reconcile the tracker, refresh the dashboard, and report only exceptions. That pattern works for money, job hunt, publishing and local work alike.
Third time you do the same job the same way, save it instead of re-explaining it.
Direct answer: A workflow becomes a skill when its trigger, inputs, rules, steps, output format, QC checklist and approval gates are written down once, in testable lines, and Muse runs that text every time.
The difference is follow-through: tools, skills, scheduled jobs and finished files.
Direct answer: A chat-only assistant answers in the chat window and stops. Muse is set up to carry a job across steps: research with sources, build a finished document or page, reuse saved skills, keep standing context in memory, and run scheduled jobs, with you approving sends, payments and submissions.
Phone assistants handle quick commands; Muse handles jobs with steps and checks.
Direct answer: Built-in phone assistants are strongest at quick, single commands: timers, calls, short answers. Muse is strongest at longer jobs that need research, drafting, building, QC and follow-up over days, such as a job hunt, a book, or a tracked side income.
Share the context a job needs, keep credentials and client data out of it.
Direct answer: Give Muse the context the job needs and nothing more: never passwords, one-time codes for accounts you are not actively signing into at Muse’s request, payment card numbers, or other people’s private details. Drafts and submissions stay staged until you approve them.
Confirm a block once, stop, and reroute. Never try to bypass a bot wall.
Direct answer: If a site such as a freelance or job platform blocks automated access with a bot wall like Cloudflare, stop at the first confirmed block. Do not use bypass tools, CAPTCHA solvers or retry loops; use the platform manually, its official alerts or API, or another platform.
Your current price and limits are the ones shown in your own Muse account.
Direct answer: Muse pricing, token allowances and usage limits change during early access and can differ by plan and region, so the only reliable figures are the ones in your own account. Check there before budgeting a workflow around a limit.
An invite code pays in Muse tokens for use inside Muse, not in money.
Direct answer: A Muse invite code gives the new user, and usually the person who shared it, a token balance to spend on Muse use. It is not cash, it does not pay out, and a new user usually has to redeem it soon after joining, in Settings > Redeem token.
Start from no account at all: check your country, official signup or waitlist, create the account, then redeem a code.
Direct answer: Muse is in early access, so access starts with the official route for your region, not with an invite code and not with Muse Playbooks. Check whether your country is live, sign up or join the waitlist with your real details, create your account when you are let in, open Settings to confirm your plan, then redeem an invite code inside the app if you have one, in Settings > Redeem token, usually within 48 hours of joining.
Full walkthrough with stuck cases, on New to Muse → · Invite code GZTOI6, on Home →
Access rolls out in stages, so availability depends on where your account is.
Direct answer: Muse is in early access and availability varies by country and over time. If a feature or signup is missing for you, the reliable test is your own account and the official signup route for your region, not a list on any third-party site, including this one.
Computer use on Mac has been announced; treat it as coming, not as in your account, until you see it there.
Direct answer: Mac computer use, where Muse can operate apps on a Mac under your direction, has been announced as part of Muse’s direction. Until it appears in your own account, plan workflows that work without it, and keep approval gates on anything it would click, send or submit.
If WhatsApp appears as a connected channel in your account, you can talk to Muse there.
Direct answer: WhatsApp access is an announced channel for reaching Muse from your phone. It only works once it appears in your own Muse account; when it does, you connect your number in Settings and then message Muse in WhatsApp like a contact.
Three settings decide most of your Muse experience: what it remembers, what it can reach, and what needs your approval.
Direct answer: Set memory to the rules you want applied every run, connect only the accounts a real job needs, and keep sends, payments and submissions behind your explicit approval. Review all three whenever a lane starts or ends.
Plan, profile, notifications, connected channels and approvals — in that order.
Direct answer: Open Settings in the Muse app and work through five things: confirm your plan and account details are the ones you expect, set notifications so payment and order alerts reach you while noise stays off, check which channels and accounts are connected, review what is saved in memory, and confirm that sends, payments and submissions wait for your approval.
What you can spend inside Muse, kept separate from money you earn.
Direct answer: Your wallet or balance area in Muse shows what you can spend inside the app: your plan's token or usage balance, any invite or referral credit, and the payment method on file. Token credit extends your Muse use; it is not cash, it does not pay out, and it never counts as hustle income.
Redeem promptly after joining — the usual window is 48 hours.
Direct answer: To redeem an invite or token code, join Muse first, then go to Settings > Redeem token in the app, enter the code exactly as shown, and confirm the token balance landed before starting a large job. A code is credit for Muse use, not cash.
Announcements describe direction; your account decides what you can use today.
Direct answer: Connect 2026 is where Meta presents what is coming for Muse and its connected ecosystem — read the official announcement at that link. Treat anything shown there as announced until it appears in your own account with working settings, and never rebuild a paying workflow around an announced feature.
Each capability gets the money edge, who it helps, and a prompt to try it.
Direct answer: Muse Playbooks tracks Muse itself in News: when a capability appears, it is written up with what it is, who it helps, the money edge, what to watch out for, and a prompt, and it is labeled by status so announced features are never presented as live ones.
Library
A skill is a saved set of instructions Muse follows every run, so your rules survive between sessions. Every skill below is opened up in full: what goes in, the steps it runs, the QC it must pass, what comes out, and how it fails.
Ranks roles against your resume and floor, drafts applications under a no-guessing rule, preps interviews from your real work.
Stale listings copied across boards; portals with no guest / Google / LinkedIn sign-in that burn an hour for one application; headline salary that loses once match, health costs and PTO are counted. The skill’s rule: verify at source, compare total pay, skip broken funnels unless the match and pay are exceptional.
Prompt matrix, 100% zoom QC notes, and the upload-ready CSV for every batch.
Duplicate-feeling batches that cannibalize each other; artifact-heavy images that earn rejections; a subject category with repeated rejections or zero traction — rotate it out instead of feeding it.
Natural, human-sounding local replies and promos, gated to services actually offered.
Pitching businesses that visibly do not need the service (they already have it, done well); platform service-gates blocking a reply — do not work around the gate from another profile; copy that sounds like a template, which neighbors ignore.
Catches truncated text, mismatched numbers and broken formatting before anything ships.
The classic: only the first wrapped line of a paragraph renders, and the rest silently vanishes. Also stale numbers copied from an earlier draft, and “looks fine in the editor” formatting that breaks in the exported file. The rule is absolute: QC the artifact, not the intention.
The 10-minute sales-rank test that decides pass / fail before you build a book.
Themes that feel charming and sell nothing; one outlier hit surrounded by dead titles (that is luck, not a niche); character-based books where the character drifts between illustrations — a structural problem no prompt polish fully fixes. Validate the niche first, then validate the format.
Everyday video and audio jobs with the exact commands, explained in plain language.
Re-encoding when a stream copy would do (slow, quality loss); silent outputs from a wrong audio-stream pick; gigantic files from skipping a resolution check. Inspect first, convert once.
Logs every payment and expense as it happens, holds the set-aside, and reconciles on a schedule.
Reconstructing a year of records in April; spending gross and meeting net at filing time; one mixed ledger for every business and personal spending; a sweep cron still running for a lane that was killed months ago.
Money
Typical ranges, not promises. Most side hustles earn far less than headlines suggest — a realistic median is roughly $200–$400 a month — so price for fees and taxes from day one, and pick one lane.
| Hustle | Startup cost | Time to first $ | Typical pay | Best for | Start |
|---|---|---|---|---|---|
| Job hunt (better role) | $0 | 4–12 weeks | Often +$10k–$30k/yr vs. current | Anyone underpaid in their current role | Starter prompt → |
| Freelance docs / SOPs | $0–$50 | 1–4 weeks | $30–$75/hr skilled | Writers, trainers, ops people | Starter prompt → |
| Virtual assistant work | $0 | 1–3 weeks | $15–$30/hr | Organized generalists | Starter prompt → |
| Local services | $0–$200 tools | Days–2 weeks | $50–$150/job typical | Hands-on skills, fastest local cash | Starter prompt → |
| KDP puzzle / activity books | $0–$30 | 1–3 months | ~$2–$4 royalty per sale | Patient builders, validated niches only | Starter prompt → |
| Adobe Stock | $0 | 1–6 months | Cents–dollars per download, compounding | Background income, weekly habit | Starter prompt → |
Ranges are general information drawn from common market rates, not a guarantee of earnings.
The stuff most hustle content skips — and the stuff that actually decides whether a lane survives.
Separate account, tax set-aside, expenses tracked, budget on your lowest month.
Side income is taxed on top of a paycheck, so it lands in your highest bracket. Working system: keep side income in a separate account, move roughly 30% of profit aside for taxes as it arrives, track every expense and business mile when it happens (not in April), and budget the lane on your lowest month, not your best. What breaks people is not the tax rate — it is spending gross and meeting net later.
A portfolio of real work beats claims. One credible cert for your lane, not ten.
Employers and clients pay for proof: samples of real work (redacted where needed), measurable results, and references. Build three strong samples in your lane’s real format, keep their original professional formatting — the formatting itself is evidence — then add one credible certification that the lane actually respects. Ten half-finished certs signal dabbling; proof plus one cert signals hireable.
Quote before work, deposit up front, scope in writing, negotiate the whole deal.
Never start unquoted. A workable pattern: flat quote in writing, a modest deposit before work begins, parts or extras agreed before you buy anything, and any sourcing fee agreed in advance — never a surprise. When negotiating a salary or rate, negotiate total pay: base, match, health costs, PTO, timeline and revisions. The headline number is the easiest one to move and often the least valuable.
Your first 10 jobs buy ratings and proof. Ask once, save everything.
Early jobs are marketing spend disguised as work: price fairly, deliver slightly more than promised, and finish cleanly. Then ask once for a review — one message, both review links — and never chase a second ask, especially with someone who already reviewed you. Save every result, photo (with permission, no personal details) and kind word as a case study; that file is what your next ten clients are actually buying.
Match timing, health costs, open enrollment, PTO — quitting at the wrong week is expensive.
Before changing jobs, run the calendar: employer retirement match vesting and eligibility dates, whether you must be employed on the last day of a plan year to receive a match, open enrollment windows, health-plan costs on both sides, tobacco or wellness surcharges and how their refund rules actually trigger, and PTO payout rules. A raise that lands one week too early or too late can quietly cost more than it gains. Compare total pay across a full year, not the first paycheck.
Cap a new lane at 5–10 hours a week. Automate anything you do three times.
One lane until it pays. Cap it at 5–10 hours a week to start so it survives a bad work week. Anything you do three times gets a template, a checklist, or a skill — that is exactly what Muse skills are for. Keep a simple weekly log: shipped, failed, one number, one lesson. If a tool, site or portal is confirmed blocked, log it once and reroute; do not pay rent to a dead route with daily retries.
No real numbers, no fees shown, no time to first dollar? Skip it.
Three questions end most hype: What are the real numbers? (not testimonials — medians). What are all the fees? (platform, payment, materials, your hours). When is the first dollar, in weeks? If a pitch cannot answer all three, it is selling the pitch. And remember: “low competition” with no buyers is not an opportunity, and rewards paid in tokens or points are not cash until they convert.
One small, finishable move per day.
Score any idea on three lines before you start it.
Score any idea on three lines before you start: Expected dollars (realistic, not best-case) · Time to first dollar · Effort cost (hours/week, ongoing). Then write the kill rule down. If you can't fill in line one with a real number, it's a hobby — enjoy it as one, don't budget on it.
Field notes
Daily News first, newest first. Then two tracks: Muse Updates and Discoveries, lessons from real builds.
Daily News
Last updated: Tuesday, November 3, 2026 · 6:00 AM ET — daily update. Dates below are the Muse Playbooks coverage dates for this November 2026 snapshot. Open the source link on any item to check the original. Every summary is in Muse Playbooks's own wording.
When a fact changes — whether it comes from Muse news, world news, or progress in a lane — Muse Playbooks updates every area that mentions it, not just this News page. Guides, Tools, Tasks, Playbooks, New to Muse, the Home FAQ, and Crons & Dashboards must all say the same thing about pricing, availability, features, and steps, with the same dates and the same source links. If you see a Daily News item change a price, a country status, a feature status (Live / Announced / Reported), or a step, you should find that same change, with that same date and source, everywhere else on this site that covers it. If two pages disagree, the newer dated source wins, and the older page is the one that is wrong — treat your own Muse account as the final word on pricing, availability, and features.
A November 3 recap of Meta's October 21 earnings call reports that more than 10 million people in the US and Canada set up Muse agents in about six weeks, with roughly 25% using Muse daily. Shopping is reported as the leading use case, with average purchases over $200, and monetization is planned through transaction fees, premium subscriptions, and enterprise offerings including Muse Code and business agents.
Source: Zacks — Meta Q3 earnings call recap, November 3, 2026
How this affects a Muse user: Shopping and business tasks are where Meta is investing first, so expect those workflows to get the most attention.
A November 3 comparison places Meta's Muse alongside Anthropic's Claude Duet and OpenAI's Dots. It notes Muse launched in September as a free agent with usage limits, reports Power ($20/month) and Maximum ($100/month) tiers for heavier use, and reports Singapore was added October 27. Check any price or country status in your own Muse account before relying on it.
Source: Monolith Design — Claude Duet vs OpenAI Dots vs Meta Muse, November 3, 2026
How this affects a Muse user: Use the comparison to judge whether Muse's free tier covers your tasks before paying for a higher tier.
Meta's Muse bug bounty program reportedly paid about $1.2 million across 64 valid findings from more than 630 reports, including $300,000 for an authentication flaw found before launch, with a reported $130,000 ceiling for a prompt-injection attack against one user.
Source: Runtime Wire — Meta Muse bug bounties, November 2, 2026
How this affects a Muse user: Agent access to email, accounts, and purchases makes security fixes directly relevant, so keep approvals per action and review connected accounts.
Coverage of November 1 reports says JPMorgan Chase, Charles Schwab, Morgan Stanley, Amazon, Neiman Marcus, and Kohl's moved to restrict or block outside AI agents, with Amazon reportedly asking Meta to remove it from the Muse shopping experience, while Walmart, Shopify, and Best Buy chose partnerships and dedicated gateways.
Source: FTMQ Tech — Amazon and JPMorgan move to block Meta's Muse AI agent, November 2, 2026
How this affects a Muse user: Some stores and banks may not work through Muse, so check the merchant directly if a shopping or account task fails.
Change Agents announced November 2 that it joined Meta's Muse Developer Program to evaluate connecting its Catch-Up and Beacon apps to Muse. The company states any future availability is subject to development, testing, and approval, with no assurance of launch.
Source: Morningstar / GlobeNewswire — Change Agents joins Meta's Muse Developer Program, November 2, 2026
How this affects a Muse user: More third-party apps may become reachable through Muse, but only after they clear approval, so do not plan a workflow around an unlaunched integration.
A November 2 report says US Senators Ron Wyden and Edward Markey are investigating privacy and safety questions about Meta's AI glasses, including accidental camera activation, tampering with the recording indicator, and face and license-plate handling, and asked Meta to respond by November 26. This is a request for answers, not a legal finding.
Source: Neoteo — Wyden and Markey reportedly asked Meta about AI glasses privacy, November 2, 2026
How this affects a Muse user: If you use Muse on glasses, watch the privacy settings for visual data from AI experiences while this inquiry is pending.
In Muse Playbooks's reading of the test, the pattern matters more than any single score: Muse was strongest when the job had clear source material to work from, like preparing a meeting from the right thread, and weakest where one unchecked detail could slip into a tidy-looking summary. The tester also had a brief finish after the app was closed.
Source: Independent tester — 10 business tasks
How this affects a Muse user: Give Muse the source, not just the topic, and keep a QC pass on anything that will be sent or filed. That same rule is stated the same way in New to Muse, the Playbooks, and the Guides — see the coherence note above.
Meta's own recap is the source for what was announced for Muse and its connected products. Muse Playbooks treats everything in an announcement as Announced until it shows up in a user's own account and runs once — that label is used consistently in Guides, News, and New to Muse.
Source: Meta — Connect 2026, everything we announced (official)
How this affects a Muse user: Do not rebuild a paying workflow around an announced feature. Keep working on what is Live in your account today, and re-check Settings after announcements.
A creator test in late October, covered here on November 2, connected social accounts and asked Muse where growth was stalling and what the account's own numbers suggested posting next. It is a Reported / Tested creator use, not a Meta promise about results.
Source: Creator Instagram and Threads analysis test
How this affects a Muse user: If you post for a lane, let Muse mine your own past posts first — captions and ideas built on what already performed beat generic ideas. The Playbooks' Verified in the wild sections use this same source and label.
Beyond Muse itself, the broader shift Muse Playbooks is watching is agents moving from one-off chat to standing jobs: watching an inbox for the enquiries that pay, and replacing daily dashboard-checking with one weekly briefing a person actually reads.
Source: Creator social media and brand deal tracking
How this affects a Muse user: Copy the pattern, not the hype: one weekly briefing and one tracker beat daily checking. That is exactly how Crons & Dashboards on this site is set up, with the same dates and labels as here.
Two pilot stories Muse Playbooks tracks: a farm operator using Muse as general office help so more hours go to the farm itself, and a grooming company that had Muse work through a very large set of customer reviews to surface strengths and weaknesses. Both are owner-reported through Meta's pilot coverage, labeled Announced / Reported on the Playbooks, not independently audited.
Source: Small business pilots — Bennett Orchards and Live Bearded (US Chamber)
How this affects a Muse user: The safest first handover is admin around the job — reviews, summaries, follow-ups — not the job itself. Start there, log it in a tracker, and judge it after a week.
In the second pilot write-up Muse Playbooks tracks, a grocery owner working long weeks describes Muse taking on roles only they used to cover, and a boutique owner describes an audit of email automations that found duplicate flows and rebuilt them with tracking.
Source: Small business pilots — grocery and boutique (Small Biz Trends)
How this affects a Muse user: An audit-and-rebuild of one messy flow — email, follow-ups, or a ledger — is a concrete, low-risk first project with a clear before and after.
Across the tests Muse Playbooks follows, the dividing line for agents in general is follow-through: chained steps, a finished file or brief, and a check the user can inspect. Answers that sound right but cannot be verified against a source are failing that test everywhere, not just in Muse.
Source: Independent tester — 10 business tasks
How this affects a Muse user: Judge every Muse run the same way: finished output, opened and read, numbers traced to a source. If it cannot pass that, treat it as a draft, not a result.
Muse Playbooks's standing field note on automated access hitting a bot wall: confirm once like a normal user, log it once, and reroute to manual use, official alerts or an API, or another platform.
Source: Discoveries — When a platform blocks you, on this page
How this affects a Muse user: Never use bypass tools or retry loops. The same stop rule appears, worded the same way, in New to Muse troubleshooting, Guides (Safety), and the Playbooks.
Generated documents can report the right page count while truncating text. Muse Playbooks's QC habit — export, open, read, especially instruction pages — comes from this lesson.
Source: Discoveries — Render it, then read the actual page, on this page
How this affects a Muse user: Build the QC step into the prompt itself, every time. Tasks, Tools, and every Playbook on this site do it the same way.
An empty niche and a generous token offer can both look like opportunity. Muse Playbooks's rule is to price both in dollars and weeks before spending time on either.
Source: Discoveries — Low competition can mean low demand, on this page
How this affects a Muse user: Count invite code GZTOI6 as working token credit in Settings > Redeem token, never as income in a hustle plan. That wording is the same on Home and in Guides (Cost).
Track 1 · Muse itself
Muse evolves in early access, so always check the capability in your own account before building a workflow on it.
The single biggest time saver for repeat work.
Hand Muse a whole job, not one question at a time.
Your work can become a page other people can open — like this site.
Demand checks, competitor scans and price checks, with the receipts attached.
Guides, SOPs, spreadsheets and CSVs built to a spec, not pasted from chat.
Log it once in the ledger. Add a dashboard to read it, and crons to reconcile it.
Read the official announcement, then check your own account before planning around any of it.
Track 2 · Field notes
Most buyers meet your book as a 200-pixel thumbnail in search results. Covers designed at full size — small elegant type, subtle imagery — disappear at that scale. What works: a bold sans-serif title that survives shrinking, one clear theme image, high contrast, and the badges buyers scan for (puzzle count, “Large Print”) placed where a thumbnail keeps them visible. Always test your cover at thumbnail size before publishing. If you can't read it small, neither can a buyer.
The cheapest research is reading what's already selling. Sales-rank data on existing titles tells you whether a niche has buyers: when most top titles in a theme rank poorly, that theme doesn't sell, no matter how good your version would be. Ten minutes of rank-checking has saved more wasted build days than any productivity trick. Validate first, build second — in publishing, in stock media, in local services, everywhere.
Generated files lie politely. A document can report the right page count while quietly truncating a sentence on page 2. The fix is boring and total: export the real file, open it, and read it — especially instruction pages, disclaimers and anything auto-wrapped. Every Muse Playbooks playbook carries a QC step for exactly this reason. Tools that generate need humans (or a second pass) that verify.
Automation fails in two ways: stopping too early, and grinding forever against a wall — a blocked site, a closed portal, a route that no longer exists. When a block is confirmed (not guessed), the productive move is to stop, log it, and reroute your hours. Retrying a dead route daily isn't persistence; it's a subscription to frustration. Note the block once, clearly, and move on.
What it looks like. A "Verify you are human" or "Checking your browser" screen that never clears. A CAPTCHA that loops no matter how many times you solve it. 403 or 429 errors on pages that open fine for everyone else. Or the quiet version: you can sign in, but messages, searches or applications silently fail or never update. If you rely on automated access to a freelance or job platform, this is the failure you will eventually meet.
Why it happens. Most large platforms sit behind bot management — Cloudflare is the common one — that scores every visit: IP reputation (shared, VPN and data-center ranges score badly), browser fingerprint, request speed and pattern, and signals that a browser is automated rather than hand-driven. Advanced users trip it precisely because their setup looks like bot traffic: scripted page loads, rapid repeated checks, unusual browser fingerprints. Many platforms also prohibit automated access in their terms, so the wall is policy, not just a glitch.
What to do. Stop at the first confirmed block. Confirm it once, yourself, in a normal browser the way a regular user would — if the platform works there, the wall is aimed at the automated route, not at you. If it blocks normal human use too, use the platform's official support channel. Then stop. Do not try to bypass it: no CAPTCHA-solving services, no fingerprint spoofing, no rotating proxies, no retry loops. A bypass attempt can escalate a soft block into an account flag or a ban — and losing the account costs far more than losing the route. Hammering a wall also burns the one resource you cannot refill: your hours.
Park the route, use a lane that is open. Log the block once, with the date, and mark the route parked in your tracker so you stop paying it daily attention. Then move the same work to a channel that is open: do it manually in the platform's own website or app, turn on the platform's official email alerts and saved-search notifications instead of scraping for updates, use an official API if the platform offers one, or take the lane to a different platform. Keep your proposals, portfolio and tracker in your own files, so a parked route never strands the business that ran on it.
What it means for advanced users. Any workflow built on automated access to someone else's platform is rented, not owned — it can end on any Tuesday, without notice, and arguing with the wall will not bring it back. Build accordingly: every automated step gets a manual fallback you have actually tested, your records live outside the platform, and no single route carries a whole income lane. The durable edge was never the automation itself. It is the process, the proof and the QC — and those still work, by hand, on the lanes that are open.
New platforms always look like open fields: few guides, few creators, little competition. Sometimes that's an opportunity. Often it's a signal that the audience is still small. The same caution applies to rewards: referral programs that pay in tokens, credits or points can be genuinely useful, but they aren't income until they convert to money. Evaluate every “opportunity” in dollars and weeks, not in vibes.
Money operations
Tracking taxes and expenses is a job Muse does on its own — no dashboard required. Keep one ledger per business, log every payment and expense the day it happens, hold a tax set-aside on every payment, and close the month against real payout records. That ledger is the whole system on this page.
You can also build a dashboard from the ledger to read it faster — the tax dashboard in the second section below. It is one example of a dashboard, not the only one you can build. Dashboards and crons in general, plus the full list of dashboard examples for every lane, live on Crons & Dashboards →
Every payment and expense logged the day it happens, with the set-aside moved before the money gets spent.
Rebuilding a year of records from memory in April. Spending gross and meeting net at filing time. Mixing two businesses, or business and personal, in one undifferentiated list.
The tax dashboard — one page that answers two questions: am I ahead on taxes, and is the lane paying?
The ledger works on its own. A dashboard just makes it readable in seconds instead of a scroll through rows. This is the setup for the tax dashboard specifically — it appears in the examples list on Crons & Dashboards, alongside the job hunt, publishing and local services dashboards, each linking to its own setup.
Panels that earn their placeFive dashboards, none of them updated. Pretty charts fed by hand-typed numbers. Business and personal panels mixed together, so neither view gets read.
A cron is a job Muse runs on a timer, whether or not you open a chat.
These are the crons specific to tax and expense tracking. What crons are in general, and the ones for other lanes, are covered on Crons & Dashboards.
Set the schedule once. The job runs, writes its result to the ledger or the dashboard, and only pings you when something needs a decision.
| Job | Schedule | What it does |
|---|---|---|
| Expense sweep | Weekly, e.g. Sunday evening | Reconciles the ledger against payments, receipts and notes since the last run. Flags anything ambiguous for your call instead of guessing a category. |
| Set-aside check | Every 2 weeks | Compares set-aside held against set-aside owed on logged profit. Alerts only when it is short. |
| Payment alerts | Daily digest, or as they land | Reports new orders and payments so they get logged the same day. Everything else stays silent. |
| Deadline watch | Weekly | Surfaces upcoming quarterly estimate dates, renewals and filing deadlines you chose to track, early enough to act on. |
The ledger, the tax dashboard built from it, and the crons that keep both current — in one setup run.
Advanced
Replace first: your income lanes, set-aside percentage, expense categories, and the tax deadlines you track.
Tracking system
The system that keeps any lane visible between sessions: a tracker or ledger underneath, one pinned dashboard on top of it, and crons — scheduled jobs — that keep both current and flag only what needs you.
The source of truth. One row per application, job, book, batch or payment, with status, next step and date. If it is not in here, it did not happen — a dashboard cannot show work nobody logged.
One page, built from the tracker, pinned as the surface you open first. Totals, open items, deadlines and lane status at a glance. It holds no hand-typed numbers of its own, and it shows its last-reconciled date.
Scheduled jobs that run on a timer: reconcile the tracker, update the dashboard, watch deadlines, and report only exceptions. Crons log, draft and alert. Sends, payments and submissions still wait for your explicit go.
A dashboard is not one thing — you build one per lane, from whatever that lane tracks. The tax dashboard is just the first example below, and the one most people build first. Every example links to the content that sets it up.
| Example dashboard | Source underneath | Panels that earn their place | How to set it up |
|---|---|---|---|
| Tax & expense dashboard | Income & expense ledger, per business | Income by lane, set-aside owed vs held, expenses by category, uncategorized items, next tax deadline | First set up the ledger itself in Money Ops — the dashboard build steps are in that page's Example Dashboard section. Kept current by a weekly expense sweep cron + biweekly set-aside check. |
| Job hunt dashboard | Application tracker (company, role, date, status, next step) | Live applications, interviews coming up, follow-ups due, closing dates, replies waiting on you | The tracker fields and tracking setup are in the Job Hunt playbook. Kept current by a weekly deadline & follow-up watch cron. |
| KDP & Adobe Stock dashboard | Pipeline tracker (niches tested, books / batches, QC result, sales or downloads) | In production, passed / failed validation, shipped this month, best and dead categories | Tracking setup is in the KDP playbook and the Adobe Stock playbook. Kept current by a weekly batch & pipeline cron, same day each week. |
| Local services & freelance dashboard | Jobs tracker (inquiry, quote, deposit, booked, done, reviewed) | Quotes waiting on a reply, jobs booked this week, deposits held, finished jobs with no review ask yet | The jobs tracker and its tracking setup are in the Local Services playbook. Kept current by a daily payment / order alert + weekly jobs sweep cron. |
| Cron | Schedule | What it does | Reports |
|---|---|---|---|
| Tracker sweep | Weekly | Reconciles the tracker against chats, mail and notes since the last run. Adds missed rows, flags duplicates. | Only gaps and ambiguous items, for your call |
| Deadline watch | Weekly | Surfaces closing dates, follow-ups due, renewals and filing deadlines early enough to act. | Anything due in the next 7 days |
| Payment & order alerts | As they land, or daily digest | Surfaces new payments and orders so they get logged the same day. | Payments and orders only — everything else stays silent |
| Dashboard refresh | Weekly, after the sweep | Rebuilds dashboard totals from the tracker and stamps the new reconciled date. | Any total that changed, and any panel now stale |
| Lane review | Weekly or biweekly | Rolls up shipped, failed, dollars and hours per lane against its kill rule. | One keep / fix / kill recommendation per lane, with numbers |
Pick one example from the list above and build only that. The tax dashboard is the worked example most people start with — Money Ops walks through the ledger underneath it and the dashboard build. Once one dashboard runs clean for a month, copy the pattern to another lane.
Stale panels, hand-typed totals, ghost crons, and dashboards nobody opens.
Skim page · No dropdowns
This page has no panels to open. Read top to bottom and you have the whole start: access, your first task, your numbers, tracking, and what to do when you are stuck. Everything linked here says the same thing, with the same dates and sources — that is the coherence rule stated on News.
Muse is Meta's personal AI agent. You give it a job and the context it needs, it works through the steps — research, draft, build, check — and brings back a result for you to approve. It is for jobs with steps, not just quick questions. You approve anything that sends, pays, or submits. Approval is per action.
Check if Muse is live in your country in the Availability Checker in Tools (snapshot: November 2026 — your own account and the official signup for your region are the final word). Use only the official signup or official app listing for your region, with your real details. If signup is open, create your account and finish its confirmation. If you get a waitlist, join once and watch that email. If your country is Not yet, join the waitlist once and re-check after announcements — do not use a VPN or a borrowed address to fake a region. Once you are in, open Settings, confirm your email, time zone, plan and limits, then redeem an invite code straight away in Settings > Redeem token. If you want to use the Muse Playbooks code, it is GZTOI6 — a Muse invite/token code, not cash, usually redeemed within 48 hours of joining. An invite code does not create an account; you join first, then redeem.
Minutes 0–2: confirm Settings — email, time zone, plan, balance — and redeem a code if you have one. Minutes 2–4: give Muse one standing rule (your floor, your hours, or “drafts wait for my approval”) and ask it to repeat back exactly what it saved. Minutes 4–6: pick one small job you already need, with a one-sentence definition of done. Minutes 6–11: run it with your numbers and source material, and end the prompt with a QC instruction: check every name, date and number against the source, flag anything guessed. Minutes 11–13: read the result against the source yourself, open any file it made, fix one thing. Minutes 13–15: log one tracker line — date, job, result, keep / fix / kill — and save the prompt if it worked.
A task is one job with a prompt. A playbook is a tested method for a repeated job: goal, setup numbers, numbered steps, a QC check, and a kill rule. A skill is that playbook saved as instructions Muse reuses every run. A tracker or ledger is one row per item — application, job, book, batch, payment — with status and next step. A dashboard is one pinned page built from that tracker, with its last-reconciled date at the top. A cron is a scheduled job that reconciles the tracker, refreshes the dashboard, and reports only exceptions. Crons draft, log and alert; they never send, pay or submit.
Write down three numbers: expected dollars (realistic, not best case), time to first dollar, and hours per week you can honestly give it. Add your pay or price floor, and a kill rule — the result after which you stop that lane. If you cannot put a real number on expected dollars, treat it as a hobby, not a plan. Compare lanes on the Hustles page.
Pick one lane only. Job Hunt: floor first, verify on the employer's own page, QC before submit, never guess a screening answer. KDP Puzzle Books: pass the sales-rank test before building, solver-check every puzzle, test the cover at thumbnail size. Adobe Stock: small weekly batches, QC at 100% zoom, build the CSV as you go. Local Services: quote and deposit in writing before work, one review ask after a happy job, never a second ask to someone who already reviewed you. Full steps are in Playbooks, each with a Verified in the wild section labeled Tested, Reported, or Announced by Meta, with source links.
Start the tracker the same day the work happens. Build the dashboard from the tracker only — no hand-typed totals. Pin it. Set one weekly cron to reconcile it and report only gaps, overdue next steps, and deadlines in the next 7 days. For money, keep one ledger per business, log every payment and expense the day it happens, and move the tax set-aside — Muse Playbooks's working rule is 30% of profit — before the rest is touched. Full setup: Crons & Dashboards and Track Taxes & Expenses.
Day 1: one small real job, end to end. Day 2: save your standing rules and confirm each one back. Day 3: connect only what that day's job needs, and test the connection. Day 4: start the tracker. Day 5: build and pin the dashboard. Day 6: set one cron. Day 7: review the dashboard and decide keep, fix, or kill from the numbers. There is a tick-off version of this list on New to Muse.
Cannot sign up: you are likely Not yet or waitlisted — waitlist once, real details, re-check after announcements. Code will not redeem: check spelling, the 48-hour window, and whether it was already used on this account, then official support. Confident wrong answer: ask for sources, open the source, re-run with a verify-and-flag-guesses instruction. Generic output: add numbers, source material, format, and a done definition. A feature is missing: it is Announced, not Live for your account yet — keep the manual way. A site shows a bot wall: stop, log it once, reroute — no bypass tools. Dashboard is stale: reconcile today and set the weekly sweep. Tokens draining: size the job in the Token Estimator first and run one job per prompt.
Pricing, token allowances, availability by country, and feature status change in early access and differ by plan and region. The figures in your own Muse account are the source. Everywhere on Muse Playbooks — this page, Home FAQ, New to Muse, Guides, Tools, Tasks, Playbooks, and Crons & Dashboards — the same fact carries the same date and the same source link, and when Daily News updates a fact, the other pages update with it. That is the coherence rule, stated in full on News.
