LinkedIn maxxing with Muse: 12 prompts for an agent that keeps working after you close the app

LinkedIn maxxing with Muse means giving the agent your business context once, saving it as a document and a skill, and then running 12 prompts that research topics, draft posts, review them and report on results. Muse has its own cloud computer, remembers what you tell it and keeps working after you close the app, so it suits a weekly content routine; publishing stays with you.

Asking any AI for "a LinkedIn post about X" skips the parts that make a post worth reading: knowing your customers, choosing an angle, collecting proof and knowing what you can honestly claim. The prompts below turn that groundwork into a routine you can repeat, from market research and a content calendar to spotting which posts led to real business conversations.

What you end up with:

• A voice guide based on how you really communicate.
• A research habit that finds ideas before you need them.
• A drafting and review loop with clear quality standards.
• A reporting loop that helps you choose the next topic.

Muse fits this kind of setup because it is an agent with a cloud machine of its own. It keeps memory, files and Artifacts between conversations and carries on working in the background. Keep this page for your next planning session.

What should Muse know before you run the first prompt?

"Sound human" plus a job title gives Muse almost nothing to work with. It needs to understand who you sell to, what you believe and which claims you can prove.

Send this first:

This is the brief for all my LinkedIn writing. Save it as a reference document and tell me where it lives.

• The company: [name, website, product or service, typical sales cycle].
• The people I write for: [roles, company sizes, industries, locations]; their problems: [problems]; their usual objections: [objections]; the moments that send them looking for help: [triggers].
• How we position ourselves: we help [audience] get [outcome] through [approach]; our clearest differences are [differences]; we are a poor match for [disqualifiers].
• My views: [three opinions about my industry]; I push back on [common advice] because [reason].
• Proof you may use: [approved case studies, results, examples, personal experience]. Never make up customers, revenue, stories or quotes.
• My voice: [10-20 posts] and [transcripts or writing samples]; avoid [phrases, tones, formatting habits].
• My goal on LinkedIn: [qualified conversations / deal support / hiring / awareness]; my offer: [offer]; the step I want interested readers to take: [action].
• Ground rules: use only connected sources you are allowed to access; if a source is out of reach, ask me for an export or a paste; leave publishing, comments, invitations and messages on LinkedIn to me; write drafts into [review destination], keep evidence and inference separate, and flag anything that needs my input.

In Muse: make this the first message of a new chat in the Muse app or at muse.ai, and attach your sample posts and transcripts as files instead of pasting them. For [review destination], name a document Artifact, such as "LinkedIn review doc". Muse can create documents, trackers and dashboards as Artifacts and keep them on its VM, so every later prompt writes into the same file.

Revise the brief whenever your offer, audience or views change.

Where does Muse store this context?

Muse holds context in three places, and the setup works best when you use all of them:

• Memory. Muse remembers details you mention and brings them back later. You can open the memory files, correct them and tell Muse to forget anything out of date. Read them after the first week: a wrong or overshared memory quietly colours every draft.
• A document on its VM. Files, Artifacts and memory all sit on your own Muse Secure VM. Ask Muse to save the brief as a document called "LinkedIn context" and to read it before every content request.
• A skill. A skill is a saved set of instructions Muse loads when a request matches it. Ask Muse to store your ground rules as a skill named "LinkedIn content": read the context file first, draft only, never publish, flag missing evidence. The rules then apply in every new chat without another paste.

Once the context is saved, set up the skill with one message:

Create a skill called "LinkedIn content" from my LinkedIn ground rules. For any LinkedIn request, first read the "LinkedIn context" and "Voice guide" documents. Only draft. Never publish, comment, connect or send messages for me. Mark any claim that has no source in my files.

When the routine is running, each piece lives here:

Open a side chat for each job (research, drafting, review, reporting) so one long thread doesn't muddle the context. If you use Muse Code, the terminal agent, the equivalent is an AGENTS.md file at the project root (muse init creates it) plus Markdown notes in .agents/memory/ with a MEMORY.md index.

Part 1: research and writing (prompts 1-8)

Prompt 1: how do you get Muse to build your voice guide?

A voice guide is useful when it explains your writing decisions: how you introduce an argument, where you bring in examples and how you link one idea to the next. "Conversational and authoritative" tells a writer nothing they can act on.

Study the posts and transcripts I've uploaded and write me a usable LinkedIn voice guide. Cover:

• how I begin posts, explain ideas and end an argument;
• my usual vocabulary, sentence length, paragraph length and how often I use lists;
• how I sound when disagreeing, unsure, funny or confident;
• phrases that come naturally to me, quoted word for word with where they came from;
• habits that show up only in my weaker pieces and should not become rules;
• a list of things to avoid, including needless repetition, made-up drama, unsupported certainty and too many one-line paragraphs;
• three before-and-after edits that move generic copy into my voice without changing its meaning.

Make it specific enough that someone else could write with it.

In Muse: Muse has no LinkedIn connector and can't read your profile. Export or copy your posts into a file and upload it; if your transcripts are in Google Drive, switch on the Google Workspace connector and name the folder. Real samples guide the model better than adjectives, so include your best posts and not only the newest. Add your own short ban list too: openers, sign-offs and filler phrases you never use. Muse, like every model, slides back to them unless you name them. Ask it to save the result as a "Voice guide" document and attach it to the LinkedIn content skill.

Why it helps: feedback becomes reusable. Instead of fixing the same phrasing each week, you improve the instructions behind every draft.

Prompt 2: how do you find customer questions worth a post?

Sales calls and customer conversations hold questions that no keyword tool shows. The useful ones usually sit next to a decision someone is struggling with.

Read the sales calls, support conversations, customer interviews and lost-deal notes from [period]. Put the questions that keep coming back into these groups:

• problems people are trying to understand;
• worries that hold up a decision;
• alternatives they compare us with;
• expectations that end in disappointment;
• questions they ask right before deciding.

For each group give an anonymised quote, its source, the number of separate conversations it appears in, the person's role and their decision stage. Offer two LinkedIn angles per group: one to help a reader diagnose the problem, one to help them weigh a solution. Rank the top 10 by how often they recur, their commercial weight and how strong our evidence is. Keep what customers actually said apart from your reading of it.

In Muse: if customer threads are in Gmail, connect it with read access only. Muse lets you choose read or send for each connector, and this prompt only reads. Upload call transcripts and lost-deal notes as files. Ask for the result as a tracker Artifact with one row per group, so prompt 6 can draw on it later. Rerun it each month on new conversations and tell Muse to update the counts in the same tracker instead of starting another.

Why it helps: you get a backlog of topics rooted in conversations your company already has, worded the way customers talk, with evidence you can verify.

Prompt 3: how do you catch topics moving from X into your market?

A debate gathering pace on X can become a strong LinkedIn post if you explain why it matters to your customers and add something of your own on top.

With your browser and public sources on X and the web, go through the past [seven days] of discussion about [topics] and from these accounts: [accounts]. Pick up to 10 developments or arguments that matter to [reader]. For each, return:

• the original source, author, date and link;
• a neutral summary of what happened or what people are arguing;
• the strongest competing interpretations;
• what it means for our audience;
• an angle we could support with our own experience or evidence.

Treat reaction counts as a snapshot, don't treat popularity as proof, and check factual claims against the original source where you can. Recommend three ideas to develop and say what I'd need to add before each could become a credible post.

In Muse: X is not in Muse's built-in connector list, so this prompt runs through its browser sub-agent on public pages. You can watch in the browser view and take over when a page asks for a login. Browsing is slower than a direct connection, so keep the account list short. Threads connects through Accounts Center, which makes it a handy second source. Ask Muse to place every link it opened next to the matching summary, so you can check a source in one click before building a post on it. To make this weekly, add: "Run this every Monday morning and update the same research document."

Why it helps: research gives you a place to start, and your judgment gives readers a reason to follow you. A good adaptation ties the debate to a choice your readers actually face.

Prompt 4: how do you study posts that beat their authors' normal numbers?

Measuring every post against a celebrity account's biggest hit warps the research. Compare each post with what its own author usually gets, then look for what explains the difference.

Here is a LinkedIn swipe file with the performance data I have: [file]. Where the data allows a fair comparison, set each post's reactions and comments against the author's median for a similar format and period. Show the maths, and flag small samples, missing numbers and posts that are very recent. For the biggest outliers, describe:

• the topic and intended reader;
• the opening, argument, evidence and ending;
• the format and rough length;
• whether the comments suggest real customer interest;
• other possible causes, such as a launch, a news event or unusually wide distribution.

Turn the findings into five reusable structures and suggest an original idea for each based on my own experience. Don't copy distinctive wording, and don't present a pattern you noticed as a rule of the algorithm.

In Muse: upload the swipe file as a spreadsheet. Muse can write and run code on its VM, so ask it to calculate the medians in code and hand back a spreadsheet Artifact with the formulas visible, not only a summary. To test your own posts against the same baseline, put their numbers into our LinkedIn engagement rate calculator.

Why it helps: you get a better set of ideas to test and avoid assuming that a copied layout will copy the results.

Prompt 5: how do you turn recordings into original posts?

A transcript can bury a strong post inside a story, a casual explanation or a disagreement that never got the attention it deserved in the original conversation.

Go through this transcript of an interview, podcast, webinar or meeting: [source]. Find eight moments that could carry an original LinkedIn post, giving preference to decisions, mistakes, surprising observations, practical explanations and opinions rooted in experience. For each moment, give:

• the exact quote and its timestamp or location;
• the main idea and who it would help;
• any context needed so the speaker isn't misrepresented;
• supporting facts we already have;
• one follow-up question that would make the idea stronger.

Choose the three best and draft a post from each, following my voice guide. Keep the speaker's level of certainty, remove confidential details and leave placeholders where a result or example still needs checking.

In Muse: upload the transcript or give Muse a link to a public recording page. Long recordings take time; Muse carries on in the background and messages you when the moments are ready. Ask for timestamps in mm:ss so you can jump to each moment and hear the tone yourself. Have it keep the three drafts in the LinkedIn review doc, outside the chat, so they reach prompt 8 unchanged.

Why it helps: you publish from experience you already have, without asking anyone to invent a fresh opinion whenever the calendar has a gap.

Prompt 6: how do you plan four weeks of LinkedIn posts?

A calendar should balance what customers need to learn with what you have evidence to say. Filling each slot with another version of your offer makes the feed repetitive.

Plan four weeks of LinkedIn posts at [three per week] from the ideas we approved in research. Include a mix of:

• practical explanations that help readers solve a problem;
• opinions backed by experience or evidence;
• customer stories we may share;
• answers to objections and decision questions;
• an occasional offer post with a relevant next step.

For each post, state the reader, the single goal, the main argument, the source material, a suggested format and the action the reader should take. Check for repeated topics, overlapping arguments, unsupported claims and too many sales posts. Write full briefs for week one and shorter outlines for weeks two to four. Leave one slot a week open for something timely; the plan doesn't need to stay fixed.

In Muse: hand this over as a goal instead of a one-off request: "Keep this calendar as my LinkedIn goal for the next four weeks." The Goals view then shows the plan and its progress. With Google Calendar connected, Muse can block writing time on the days before each slot; keep calendar changes on approval so nothing moves without you. To keep the review load light, add: "Write the first week-one brief, then stop and wait for my approval before the next." Choose slots with our best time to post on LinkedIn tool, and build any carousel a brief asks for in the LinkedIn carousel maker.

Why it helps: less weekly planning, with room left for better ideas, new evidence and market changes.

Prompt 7: how do you get drafts with a real argument?

Let the research settle the argument before Muse starts polishing sentences. Otherwise you get neat writing that spends several paragraphs on something obvious.

Write three LinkedIn drafts from these approved briefs: [briefs], using my business context and voice guide. In every draft:

• make one clear argument for one reader;
• open with a specific observation, tension, result or question the brief supports;
• walk through the reasoning with a concrete example;
• use bullets where they make parallel items easier to scan;
• vary sentence and paragraph length, with an occasional one-line paragraph for weight;
• finish with a useful consequence or a natural next step;
• aim for [180-300 words] unless the idea needs a different length.

Don't restate the same idea at the end, invent personal stories, overpromise or use formula contrasts. Add two other openings for each draft and say briefly how each changes the focus. List missing evidence separately from the copy that is ready.

In Muse: Muse is designed to keep going, so finish the message with a stop: "Do not post anything and do not open LinkedIn. Leave the drafts in the review doc and wait for me." Rewrite each first line in your own words before anything else, and if you want more openings than the two alternatives, try our LinkedIn hook generator.

Why it helps: tight constraints give you drafts that are easier to judge, so the review can focus on the argument instead of trimming filler.

Prompt 8: how do you run a separate review in Muse?

Give the reviewer a clear job: find the weak spots in each draft and explain how to fix them. A friendly "this is strong" isn't an editorial process.

Review these LinkedIn drafts against their brief, the source material, my business context and my voice guide. Score each out of 100:

• relevance to the intended reader: 25;
• specificity and evidence: 25;
• strength and clarity of the argument: 20;
• match with my voice: 20;
• readability and layout: 10.

Explain each score. Check every factual claim, mark generic passages, and flag repetition, abrupt one-line fragments and paragraphs that would read better as lists. Return a revision ready to publish, a short note on the changes that matter and any open questions. Invented evidence, leaked confidential information or a misquoted source stops the draft whatever its score. Keep everything in the review document until I approve it.

In Muse: run the review in a side chat, away from the drafting conversation, so the reviewer checks the drafts against your files instead of defending its own choices. Name the context file, the voice guide and the review doc it should use, and ask for the scores as a table at the top of the review doc so you can compare drafts quickly. Before you publish, paste the approved version into our LinkedIn post analyzer for a final check on length and structure.

Why it helps: a rubric spells out your standards. The score organises the feedback, and the source checks plus your judgment decide whether a post is ready.

How do you make the first hour after publishing count?

Prompt 8 leaves you with a revision you approve and publish on LinkedIn yourself. After that, every MAXXING method works the same way: during the first hour after the post goes live, get as many reactions as you can.

Ask the people you can reach. Colleagues, friends, family and clients all count, as long as you can contact them. Send each one the link as soon as the post is up, ask them to react, and invite a comment if they have something to add.

Without a circle like that, use a community with moderation. Pick one where real people act by hand. The groups listed below are ones we have checked and cover in other guides.

LinkedInLikes is our community for this step. It is free and needs only a public link to your post:

• Members react, comment, repost, save and follow by hand from their own LinkedIn accounts.
• Comments are written in each member's own words; you can't script them.
• If someone confirms an action they didn't do, report it within 72 hours: the points are taken back and repeat offenders are suspended.
• We never ask for your LinkedIn password, a cookie or a browser extension.

Once the post is live, add it to LinkedInLikes for comments or for likes.

Which LinkedIn pods and groups can you join?

We checked these free communities on Oct 21, 2026, and the member counts are what each platform showed that day. Each group sets its own rules through its admins, so read the pinned message before your first post.

For reactions on your own posts without searching through groups, sign up for LinkedInLikes; new accounts get 100 starter points for their own posts.

Part 2: pipeline and performance (prompts 9-12)

Prompt 9: how do you write posts about deals that got stuck?

Some of your best material may answer a worry that several prospects raised privately. Explain it well enough and the next prospect arrives better informed.

Go through the CRM records and call notes for qualified deals that got stuck in [period]. Find the objections that repeat, and keep apart:

• objections people stated in conversations;
• explanations that are still unconfirmed guesses;
• operational delays no post is going to solve.

For each repeated objection, note the person's role, the deal stage, the evidence we have and the question a helpful LinkedIn post should answer. Suggest five content briefs that each state the concern fairly, show where our approach fits, admit where it doesn't and name the proof needed. Draft the two strongest posts without naming prospects or revealing deal details. In a separate list, suggest approved existing resources a salesperson could send in a relevant conversation.

In Muse: no CRM is on the built-in connector list, so the simplest route is a CSV export of the stuck deals uploaded to the chat. Muse has no form for adding servers; you describe the connection in a chat and Muse builds a custom connector on its Secure VM. If your CRM vendor offers a way for outside tools to connect, set that connector to read-only, since this prompt never needs to change a record. Meta does not review custom connectors, so the connection is between you and your CRM vendor. Every call still goes through the Sentinel approval dialog, which is a good moment to check that Muse is reading only the period you named.

Why it helps: marketing starts answering the questions sales keeps hearing, and you stop writing persuasive posts about problems your customers don't rate as important.

Prompt 10: how do you prepare replies to people who respond?

When someone responds to a post, the right next step depends on what they wrote and whether they'd welcome a business conversation. A like by itself shows no business interest at all.

Here are LinkedIn conversations I've collected: [comments, inbound messages or interactions I recorded by hand]. Sort them into:

• questions that deserve a helpful public answer;
• conversations where a resource would help;
• clear interest in our offer;
• general discussion with no commercial next step.

For each relevant conversation, sum up the need the person stated, the business context we have and what we still don't know. Draft a reply that answers their actual message before raising anything else, and where it fits, suggest one follow-up question or an existing resource. Don't read business interest into a like, don't claim a closeness that isn't there, and don't squeeze a meeting request into a conversation that can't hold one. Leave every reply as a draft for me to check and send.

In Muse: since Muse can't read LinkedIn, paste the comments and messages into the chat or upload them as a file. Keep the replies in the chat and send them from LinkedIn yourself. If some conversations moved to email, leave Gmail on read access; Muse asks before sending anything, but a draft you copy is safer than a send you approve in a rush. For a quick first pass on a single public reply, our LinkedIn comment generator gives you options to edit.

Why it helps: replies get more thoughtful without every exchange becoming a sales sequence. Sometimes a useful answer is the whole job.

Prompt 11: how do you report LinkedIn results without overclaiming?

Impressions describe distribution. They can't tell you whether the right people became interested or whether a post helped a deal along.

Combine my LinkedIn analytics, tracked website visits, inbound enquiry records and relevant CRM data for [period]. Report:

• reach, reactions and comments by post, topic and format;
• qualified inbound conversations;
• resource sign-ups and demo requests where tracking exists;
• opportunities with a recorded LinkedIn contact;
• time spent producing and reviewing posts.

Keep directly tracked conversions, discovery people reported themselves and possible influence apart. Don't assign a whole deal to LinkedIn because one contact reacted to a post. Name the posts that reached relevant prospects, posts that reached a wide but less relevant audience and posts with too little data to judge. Close with three findings, three limitations and one experiment. Where possible, compare posts at the same age so new posts aren't measured against old ones.

In Muse: download the LinkedIn analytics export and your website report, upload both, and ask for a dashboard Artifact you can reopen next month instead of a one-off summary. Measuring every post at the same age is easier when Muse runs it on a timer: "Seven days after each post, ask me for the LinkedIn export and add that post to the dashboard." To sanity-check reach on a single post, use our LinkedIn impressions calculator.

Why it helps: you get a reason to repeat or change what you are doing, and uncertainty stays visible instead of patchy tracking turning into a confident revenue claim.

Prompt 12: how do you turn your edits into better instructions?

Each draft you edit says something about your standards. Record the lessons that repeat so you spend less time on familiar fixes.

Compare the first drafts, my edits, the published versions and any performance data from [period]. Find:

• voice corrections I keep making;
• claims that needed more evidence;
• ideas that became clearer once reframed;
• layout changes that made posts easier to read;
• topics that started useful conversations with prospects.

Propose changes to the voice guide, research briefs and editorial checklist. For each proposed rule, show an example, explain its purpose and say when an exception is fine. Keep style preferences separate from guesses about performance, and never make a permanent rule from one post that did well or badly. Give me a short change log to approve and one controlled experiment for next month.

In Muse: this prompt updates memory and the skill, so tell Muse to change nothing until you approve the change log. Afterwards, ask it to update the voice guide, the LinkedIn content skill and its memory, then open the memory files and see what it really changed. One message covers it:

Take the items I approved in the change log and apply them to the "Voice guide" document, the "LinkedIn content" skill and your memory. List each change file by file, and don't touch anything I didn't approve.

Why it helps: the routine gets easier to run over time. Corrections turn into guidance, and performance ideas stay testable instead of hardening into superstition.

How should you roll out the 12 prompts?

Add them in stages so you can check the output before taking on another recurring job.

• Week 1, the base: load your business context, run prompt 1 and compare the voice guide with writing you would really publish.
• Week 2, the research backlog: run prompts 2-4 on a manageable set of customer conversations, market sources and LinkedIn examples.
• Week 3, the first batch: use prompts 5-7 to turn approved material into a calendar and a few drafts.
• Week 4, the review: run prompt 8, publish what you approve, work the first hour on each post and note the changes you make.
• Weeks 5-6, conversations: add prompts 9 and 10 once you have enough sales context and inbound activity to make them useful.
• Monthly from then on: use prompts 11 and 12 to review results, improve the instructions and choose the next experiment.

For every recurring job, write down its inputs, where the result goes, how often it runs and who owns it. Test a run before relying on the schedule, and have Muse report missing access or unfinished work.

Muse can take over the recurring part. Describe the timing in plain words: "Every Monday at 8am, run prompt 3 and update the research document," or "On the first of each month, run prompt 11 with the files I upload." Muse works on its VM after you close the app and messages you when it needs a file or an approval. A weekly rhythm that works:

• Monday: prompt 3 runs by itself and updates the research document.
• Tuesday: you approve ideas, and Muse writes briefs (prompt 6) and drafts (prompt 7) into the review doc.
• Wednesday: prompt 8 runs in its own side chat; you publish what passes and work the first hour.
• Friday: Muse lists open questions from prompts 9 and 10 and the files it still needs from you.

Look at the Activity log after the first few runs to see what it actually did. Keep posting times in your own calendar so the agent never holds them, and publish each post yourself. In Muse Code, the same jobs can run from a script with muse exec --prompt-file.

Day one doesn't need 12 separate agents. Begin with the job that takes the most time and produces something you can judge reliably.

What does Muse still leave to you?

Muse can organise research, pull up useful evidence, prepare drafts and learn from your edits. The quality of the posts still comes from the experience and judgment you feed into it.

That means holding a real opinion, being ready to explain it, and noticing when a polished draft says something you'd never defend in a conversation.

Load your context and run prompt 1 this week; better inputs improve every step after it. When the first post goes live, add it to LinkedInLikes so it gets reactions in its first hour.

Related reading

• How to build an AI content system for LinkedIn
• LinkedIn maxxing with Codex
• LinkedIn maxxing with Claude
• The 2027 LinkedIn marketing playbook