LinkedIn Maxxing With Gemini CLI: 12 Prompts From Research to Reporting

LinkedIn maxxing with Gemini CLI means running your LinkedIn content work as files and prompts in the terminal: a business context file, a voice guide, research notes, drafts and reports that Gemini CLI reads and writes. The 12 prompts below cover research, drafting, review and reporting, and you can rerun every one of them each month.

Handing an AI a topic and asking for a post skips the part that makes a post worth reading: knowing your customers, finding an angle, gathering evidence and knowing which claims you can stand behind. These prompts turn that groundwork into a process you can repeat, from market research to a content calendar to tracking which posts led to real business conversations.

By the end you will have:

• A voice guide built from the way you really write and speak.
• A research routine that collects ideas before you need them.
• A drafting and review process with clear quality standards.
• A reporting loop that tells you what to write next.

Gemini CLI runs in your terminal, reads and writes the files in the folder where you launch it, and can be scripted. The whole system ends up as plain files and commands you can run again every month, so bookmark this page for your next planning session.

What context should Gemini CLI have before the first prompt?

A job title and a request to "sound human" give it too little. Gemini CLI needs to know who you sell to, what you believe and which claims you can back up. Paste this in first:

Treat the text below as the standing context for my LinkedIn content. Store it in a reference file I can reuse and tell me its path. Business: [company name, website, product or service, typical sales cycle]. Audience: [roles of the people who decide, company sizes, industries, locations]. Their biggest problems are [problems], the objections they raise most are [objections], and the moments that send them looking for help are [triggers]. Positioning: we help [audience] reach [outcome] by [approach]. What sets us apart: [differences]. We are the wrong choice for [disqualifiers]. My views: I hold [three opinions about my industry], and I disagree with [common advice] because [reason]. Evidence: draw only on these approved case studies, results, examples and personal experiences: [sources]. Never make up customers, revenue numbers, stories or quotes. Voice: these [10-20 posts] and [transcripts or writing samples] sound like me. Stay away from [phrases, tones, formatting habits]. Goals: my main aim on LinkedIn is [qualified conversations / deal support / recruiting / awareness]. My offer is [offer], and the next step I want interested readers to take is [action]. Rules: use connected sources only where you have authorized access; if you can't open a source, ask me for an export or pasted text. Publishing, comments, connection requests and messages on LinkedIn stay with me. Put drafts in [review destination], keep evidence apart from inference, and flag anything that needs my input.

Update this context whenever your offer, audience or opinions change.

Where does Gemini CLI store this context?

Every session loads GEMINI.md context files in layers. ~/.gemini/GEMINI.md covers everything you do; a GEMINI.md in the project folder, or in its .gemini/ folder, covers that project; and files in subfolders add narrower instructions on top.

Create a folder only for LinkedIn work, run gemini inside it and ask it to store the answers above as context/business.md. Keep the project's GEMINI.md short and import the context files instead of pasting their contents:

# LinkedIn content workspace
@./context/business.md
@./context/voice-guide.md
Save drafts in drafts/, reviews in reviews/, reports in reports/.
Do not publish, comment, connect or send messages on LinkedIn. Only prepare drafts.

Each @ line pulls a file in, so you can change the business context without editing anything else. /memory show lists exactly what is loaded, and /memory refresh reloads it after you edit a file. /memory add appends one fact to the global ~/.gemini/GEMINI.md. It suits small preferences, though users say such facts aren't always recalled, so keep anything that matters in files you control. Folders that should stay off-limits, such as contracts or billing exports, go in .geminiignore.

Settings are layered the same way as context. ~/.gemini/settings.json holds your personal defaults, .gemini/settings.json inside the folder holds the project's, and a system-wide file managed by an admin overrides both. Once all twelve prompts are in place, the folder looks roughly like this:

linkedin/
  GEMINI.md            imports the files in context/
  .geminiignore        folders the tools must not touch
  .gemini/
    settings.json      project MCP servers
    commands/linkedin/calendar.toml
    agents/reviewer.md
    skills/editorial-review/SKILL.md
  context/             business.md, voice-guide.md
  samples/  calls/  research/  drafts/  reviews/
  inbox/  replies/  reports/

A setup note: according to Google's README, on June 18, 2026 Antigravity CLI replaced Gemini CLI for unpaid-tier and Google One accounts. Gemini CLI still works with a Gemini API key, Vertex AI or an organization's Code Assist license.

Part 1: research and content production (prompts 1 to 8)

The first eight prompts take you from raw material, such as posts, calls and transcripts, to a reviewed draft that is ready to publish.

1. Build your voice guide

A voice guide earns its place when it explains the choices behind your writing: how you set up an argument, how you use examples, how you move from one idea to the next. A label like "conversational and authoritative" gives the model almost nothing to work with.

Study the posts and transcripts I've provided and turn them into a working LinkedIn voice guide. Describe: how I start posts, explain ideas and wrap up an argument; my usual vocabulary, sentence length, paragraph length and use of lists; how I show disagreement, doubt, humor and confidence; phrases I use naturally, quoted word for word with a reference to where each one appears; habits that show up in my weaker samples and should not become rules. Add a list of phrases and patterns to avoid, including needless repetition, staged drama, certainty without support and too many one-line paragraphs. End with three before-and-after edits that turn generic copy into my voice without changing what it says. Make the guide concrete enough that a different writer could follow it.

In Gemini CLI: put your sample posts in samples/ and reference the folder in the prompt as @samples/. The file tools accept more than text, since PDFs, images and audio go in as they are, so a recorded talk or a slide deck can sit beside the posts. Writing habits that apply everywhere, beyond LinkedIn, belong in the global ~/.gemini/GEMINI.md; some writers keep a second global file for creative work and swap it in when needed. Ask for the output as context/voice-guide.md. The import line in GEMINI.md already refers to that file, so after /memory refresh each new draft starts from it. /init can also generate a GEMINI.md, but it is designed for code projects; for a writing folder, the short file above is quicker to write yourself.

Why it matters: a voice guide turns feedback into something reusable. You stop fixing the same phrasing every week and change the instructions behind every draft instead.

2. Find the customer questions worth answering

Sales calls and customer conversations surface questions that no keyword tool will show you. The ones worth answering usually involve a decision someone finds hard to make.

Go through the sales calls, support threads, customer interviews and closed-lost notes I've provided for [period]. Sort the recurring questions into: problems customers are trying to understand; worries that hold up a decision; alternatives they are weighing; expectations that end in disappointment; questions they ask just before deciding. For every theme, give an anonymized quote, where it came from, how many separate conversations mention it, the role of the person asking and how far along their decision is. Propose two LinkedIn angles per theme, one that helps readers diagnose the problem and one that helps them judge a solution. Rank the top 10 by how often they come up, their commercial relevance and how strong our evidence is. Keep what customers said outright apart from your own interpretation.

In Gemini CLI: launch with gemini --approval-mode=plan. Plan Mode limits the agent to read-only research and planning until you approve, so its first pass over calls/ suggests themes without writing a file. When the grouping looks right, approve it and have the agent write research/customer-questions.md, with the file name and line beside every quote. Call recordings can go in directly as audio. Before the second pass, run /chat save calls-pass1 so you can return to the approved grouping if the write-up wanders. If a long batch of transcripts fills the session, /compress condenses the conversation so far and frees up room for the rest.

Why it matters: you get a backlog of post ideas tied to conversations your business is already having, in words your customers use and with evidence you can check.

3. Spot topics moving from X into your market

A debate picking up speed on X can become a good LinkedIn post, as long as you explain why it matters to your customers and add something beyond a rewrite.

Use Google Search grounding, web_fetch for the original pages and the public sources you can reach to review the last [seven days] of discussion about [topics] and from these accounts: [accounts]. Find up to 10 developments or arguments that matter to [customer]. For each one, give: the original source, author, date published and link; a neutral summary of what happened or what people are arguing; the strongest competing readings; what it changes for our audience day to day; a possible angle backed by our experience or evidence. Mark reaction counts as a snapshot, don't treat popularity as proof, and check factual claims against the original sources wherever you can. Recommend three ideas to develop, and say what I would have to add before each one could become a credible LinkedIn post.

In Gemini CLI: the research runs through the built-in google_web_search and web_fetch tools, and there is no direct access to X. Public posts indexed by search engines can show up, and articles and newsletters covering them usually do, but anything behind a login stays out of reach. Paste the posts you care about into research/x-notes.md, add @research/x-notes.md to the prompt, and ask at the end for the list of URLs it fetched so you can spot-check them. If you scan the same accounts each week, save the prompt as a custom command such as /research:x and pass the list of accounts as {{args}}.

Why it matters: research hands you a first lead, and your judgment gives readers a reason to follow you. A good adaptation ties the debate to a decision your readers actually face.

4. Study posts that beat their authors' usual results

Measuring every post against a celebrity account's biggest hit skews your research. Compare each post with its own author's typical numbers first, and only then look for what explains the gap.

Work through this LinkedIn swipe file and the performance data that comes with it: [file]. Where there is enough comparable data, set each post's reactions and comments against the author's median for a similar format and period. Show the math, and flag small samples, missing metrics and very recent posts. For the clearest outliers, describe: the topic and the reader it was meant for; the opening, argument, supporting evidence and ending; the format and rough length; whether the comments suggest real relevance to customers; possible confounders such as a launch, a news event or unusually wide distribution. Turn this into five reusable structural patterns, then suggest an original idea for each that draws on my own experience. Don't copy distinctive wording, and don't present a pattern you observed as a rule of the algorithm.

In Gemini CLI: reference the export as @swipe-file.csv and ask the agent to calculate the medians by writing a short script and running it with its shell tool. It asks before it runs the command, and you are left with a script you can run again instead of a number it guessed. A screenshot of a post's analytics also works, since images are accepted next to the CSV. To look at the columns yourself first, type !head -5 swipe-file.csv; that prefix runs the command in your shell without leaving the session.

Why it matters: you come away with hypotheses worth testing, without assuming that copying a winning post's layout will copy its results.

5. Turn recorded experience into original posts

A strong post can hide inside a transcript: in a story, a passing explanation, or a disagreement that deserved more airtime than it got in the original conversation.

Read this interview, podcast, webinar or meeting transcript: [source]. Find eight moments that could carry an original LinkedIn post, favoring decisions, mistakes, surprising observations, practical explanations and opinions grounded in experience. For each moment, give: the exact quote with its timestamp or location; the main idea and who it would help; any context needed so the speaker isn't misrepresented; supporting facts already on hand; one follow-up question that would make the idea stronger. Pick the three strongest moments and write a post from each in the style of my voice guide. Keep the speaker's level of certainty, strip out confidential details, and leave placeholders wherever a result or example needs confirming.

In Gemini CLI: the source can be a text transcript or the recording itself, added with @. After you approve the eight moments, save the conversation with /chat save moments. If a draft goes off track, /chat resume moments returns you to the approved list without rereading the source, and /chat share exports the session as a Markdown file for your records. For an hour-long recording, ask for a text transcript once and save it next to the audio, so later prompts reuse the text and skip processing the recording again.

Why it matters: you publish from experience you already have, and nobody has to invent a fresh opinion every time the calendar shows an empty slot.

6. Build a four-week content calendar

A good calendar weighs what customers need to understand against what you have enough evidence to say. If every slot holds another version of your offer, the feed gets repetitive.

Using the approved ideas from our research, plan a four-week LinkedIn calendar at [three posts per week]. Mix in: practical explanations that help customers solve a problem; opinions backed by experience or evidence; customer examples we have permission to share; answers to objections and decision questions; an occasional offer post with a relevant next step. For every post, set out the reader, one objective, the central argument, the source material, a suggested format and the action you want the reader to take. Check for repeated topics, overlapping arguments, claims without support and too many sales posts. Write full briefs for week one and shorter outlines for weeks two to four. Leave space each week to swap in one timely post, since the calendar shouldn't be treated as fixed.

In Gemini CLI: this is the prompt to turn into a custom command, because you'll run it every month. Save it as .gemini/commands/linkedin/calendar.toml:

description = "Four-week LinkedIn calendar from approved research"
prompt = """
Build a four-week LinkedIn calendar at {{args}} using the approved ideas in research/.
(the rest of prompt 6)
"""

Typing /linkedin:calendar three posts per week then runs it, with your posting rhythm inserted where {{args}} sits. A !{ls research/approved} block inside the prompt runs that command before sending and pastes in the current list of approved ideas, so the command never works from an outdated list; Gemini CLI asks you to confirm the shell call first. Choose the time slots with our best time to post on LinkedIn tool.

Why it matters: a calendar like this cuts weekly planning while leaving room for better ideas, new evidence and shifts in the market.

7. Draft posts that have something to say

Let the research settle the argument before Gemini CLI starts polishing sentences. Reverse that order and you get tidy prose that spends several paragraphs on an obvious idea.

Write three LinkedIn drafts from these approved briefs: [briefs], using my business context and voice guide. In each draft: make one clear argument for one intended reader; open with a concrete observation, tension, result or question that the brief supports; walk through the reasoning with a real example; use bullets where they make parallel items easier to scan; vary sentence and paragraph length, with an occasional one-line paragraph for emphasis; finish with a useful implication or a natural next step. Aim for [180-300 words] unless the idea calls for a different length. Avoid repeating the ending, made-up personal stories, inflated promises and formulaic contrasts. Give two alternative openings and explain in a sentence how each shifts the emphasis. List missing evidence separately from the copy that is ready to publish.

In Gemini CLI: ask for each draft in a separate file under drafts/ and for all evidence gaps in one drafts/gaps.md, so nothing unchecked hides inside the copy. Start the session with gemini --checkpointing: it snapshots files before each edit, and /restore rolls back a rewrite you dislike. Try more openings in the LinkedIn hook generator, tidy the line breaks with the LinkedIn text formatter and check the length with the LinkedIn character counter. If you edit in VS Code, the Gemini CLI companion extension shows proposed changes there as diffs, and you accept or reject each edit in the editor.

Why it matters: tight constraints produce drafts that are easier to judge, so your review goes into the strength of the argument and not into cutting filler.

8. Run a separate editorial review

Give the reviewer one defined job: find the weak spots in a draft and say how to fix them. A comforting "this is strong" doesn't count as an editorial process.

Review these LinkedIn drafts against the brief, the source material, the business context and the voice guide. Score each one out of 100: relevance to customers 25, specifics and supporting evidence 25, strength and clarity of the argument 20, match with my voice 20, readability and formatting 10. Show the evidence for each score. Check every factual claim, flag generic passages, and mark 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. Treat invented evidence, exposed confidential information or a misrepresented source as a blocker, whatever the score. Keep everything in the review document for a person to approve.

In Gemini CLI: the reviewer shouldn't remember writing the drafts. Either run /clear and load only the briefs, drafts and sources, or define a reviewer subagent in .gemini/agents/reviewer.md with the rubric as its instructions and call it for every batch. Store the output in reviews/. Before publishing, put the approved version through the LinkedIn post analyzer for an outside check. A third route is a skill in .gemini/skills/editorial-review/SKILL.md; skills load only when the agent activates them, so the rubric stays out of sessions that don't need it.

Why it matters: a rubric spells out your standards. The score organizes feedback, while the source checks and the editor's judgment decide whether a post is ready.

What should happen in the first hour after you publish?

When prompt 8 produces a revision you're happy with, you publish it yourself and the writing is finished. From there the playbook matches every other LinkedIn maxxing guide: the post needs as many reactions as it can get during its first hour.

Start with your own circle. The easiest people to ask for reactions are the ones you can message directly: coworkers, friends, family and clients. Send them the link as soon as the post is live.

If you have no circle to ask, use a moderated community. The free groups listed below are one option, and LinkedInLikes is another.

On LinkedInLikes, you paste the link to your new post, choose reactions or comments and set how many you want. Members open the post and act by hand from their own LinkedIn accounts, comments are written in their own words, and you spend points that you collect by supporting other members' posts. Nothing has to be installed, and we never ask for your LinkedIn password. Start from the LinkedIn comments or LinkedIn likes page.

Which free LinkedIn pod groups can you join?

These free communities were checked on Oct 21, 2026, and the member counts are the figures each platform showed that day. Every group has its own admins and rules, so read the pinned message before you share your first link.

If you'd rather skip the group chats, LinkedInLikes keeps everything on one site, with members who react by hand and a report button for actions that weren't done.

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

The last four prompts connect your posts to sales conversations and results, and feed what you learn back into the system.

9. Build content around deals that got stuck

Some of your most useful posts may answer a worry that several prospects raised in private. The aim is to explain the issue well enough that future customers arrive better informed.

Review the CRM records and call notes I've provided for qualified deals that got stuck during [period]. Find the recurring objections and separate: objections stated outright in conversations; possible explanations nobody has confirmed; operational delays that content is unlikely to fix. For each repeated objection, note the role of the person, the deal stage, the evidence available and the question a useful LinkedIn post should answer. Recommend five content briefs. Each should present the concern fairly, show where our approach fits, admit where it doesn't and name the proof required. Draft the two strongest posts without naming prospects or exposing deal details. Separately, suggest which approved resources we already have that a salesperson could share in a relevant conversation.

In Gemini CLI: if your CRM offers an MCP server, register it in the project's .gemini/settings.json and list only its read tools under includeTools, so the review can't change a record. Without an MCP server, an export added as @deals.csv works the same way. Sign-in tokens for MCP servers sit in ~/.gemini/mcp-oauth-tokens.json, outside the project, so they don't end up in a folder you share. In both cases, keep raw records in a folder the drafts never quote from, and let only anonymized briefs reach drafts/. Running the analysis with --approval-mode=plan adds another safeguard: nothing is written until you've read the plan.

Why it matters: marketing ends up answering the questions salespeople hear again and again, and you avoid writing persuasive posts about problems your customers don't care about.

10. Prepare briefs for relevant conversations

When someone responds to a post, the right next step depends on what they wrote and whether a business conversation would be welcome. A reaction or a comment on its own doesn't prove business interest.

Review the LinkedIn conversations I've pasted: [comments, inbound messages or interactions I logged by hand]. Sort them into: questions that deserve a useful public answer; conversations where sharing a resource would help; clear interest in our offer; general discussion with no commercial next step. For each relevant conversation, summarize what the person said they need, the business context available and what is still unknown. Draft a reply that answers their actual message before bringing up anything else. Where it fits, suggest one follow-up question or a resource we already have. Don't read business interest into a like, don't pretend to know the person, and don't push a meeting request into a conversation that doesn't call for one. Keep every reply as a draft for me to review and send.

In Gemini CLI: paste the comments and messages into inbox/ and add the file with @. Gemini CLI can't access LinkedIn, so replies can only end up as drafts in replies/, which is exactly what this prompt asks for. Have each reply written right under the comment it answers, so checking tone and facts takes one pass. For public replies where you'd like a few more versions to edit, the LinkedIn comment generator gives you variants to start from. When a reply is ready, /copy puts the last response on your clipboard so you can paste it into LinkedIn yourself.

Why it matters: preparation makes your replies more thoughtful without turning every exchange into a sales sequence. Sometimes the right result is simply a helpful answer.

11. Report business results with clear attribution

Impressions show how far a post traveled, but they can't tell you whether the right people got interested or whether a post helped close a deal.

Combine the LinkedIn analytics, tracked website visits, inbound inquiry records and relevant CRM data I've provided for [period]. Report on: reach, reactions and comments by post, topic and format; qualified inbound conversations; resource sign-ups and demo requests where tracking exists; opportunities with a documented LinkedIn contact; time spent producing and reviewing content. Keep directly tracked conversions, self-reported discovery and possible influence in separate groups. Don't attribute a whole deal to LinkedIn because one contact reacted to a post. Call out posts that drew relevant prospects, posts that reached a broad but less relevant audience and posts without enough data to judge. End with three findings, three limitations and one recommended experiment. Where possible, compare posts at the same age, so new posts aren't judged unfairly against older ones.

In Gemini CLI: if the exports are spread across several places, gemini --include-directories ../analytics,../crm-exports lets a single session read them all. This report is also the obvious job for headless mode: gemini -p "..." --output-format json returns structured output that a small script can save as reports/2026-11.json, which makes month-to-month comparison easy. Use stream-json when you want to follow a long run event by event. If your exports already live in a GitHub repository, the official Gemini CLI GitHub Action can run the same prompt in a workflow there. Check the reach of single posts with the LinkedIn impressions calculator.

Why it matters: better reports give you a reason to repeat or change what you're doing, and they show uncertainty openly instead of turning patchy tracking into a confident revenue claim.

12. Turn feedback into a better system

Every draft you edit holds information about your standards. Write down the lessons that keep coming back, and you'll spend less time correcting the same problems.

Compare the original drafts, my edits, the final published versions and the performance data available for [period]. Find: voice corrections I have made more than once; claims that needed more evidence; ideas that became clearer after being framed differently; formatting changes that made posts easier to read; topics that led to useful conversations with customers. Propose updates to the voice guide, the research briefs and the editorial checklist. For every proposed rule, give an example, explain what it is for and note when an exception might make sense. Keep style preferences apart from performance hypotheses. Never turn a single post's success or failure into a permanent rule. Return a short change log for me to approve and suggest one controlled experiment for next month.

In Gemini CLI: keep drafts and published versions in git and let the agent run git diff through its shell tool, so it works from your actual changes instead of its memory. With checkpointing on, each edit it makes to the voice guide can be undone with /restore, so testing a proposed rule costs nothing. Approved rules go into context/voice-guide.md and the command files in .gemini/commands/, followed by /memory refresh. Keep /memory add for one-line facts; rules you want to revisit belong in files. Once everything works, an extension can bundle your commands, context files and MCP settings, so a teammate installs the whole setup with a single gemini extensions install.

Why it matters: this is how the workflow gets easier to run over time. Your corrections become usable guidance, and performance ideas stay testable before they can harden into superstition.

How should you roll out all 12 prompts?

Build the workflow in stages so you can check the output before adding another recurring job.

• Week 1, the foundation: 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 to 4 on a manageable set of customer conversations, market sources and LinkedIn examples you've collected.
• Week 3, the first batch: use prompts 5 to 7 to turn approved material into a calendar and a few drafts.
• Week 4, the review process: run prompt 8, publish the posts you approve, work the first hour on each one and log the changes you make.
• Weeks 5 and 6, content and conversations: add prompts 9 and 10 once you have enough sales context and inbound activity to make them useful.
• Every month after that: use prompts 11 and 12 to review results, improve the instructions and pick the next experiment.

For each recurring job, write down its inputs, where the output goes, how often it runs and who owns it. Do a test run before you trust a schedule, and have Gemini CLI report any missing access or unfinished work.

Gemini CLI has no scheduler built in, so recurring runs have to be started from outside. Headless mode (gemini -p with --output-format json or stream-json) is made for scripts: a cron entry or a CI workflow can run the monthly report from prompt 11, and the official GitHub Action handles setups based on a repository. Keep every recurring prompt in a file or a custom command, so the scheduled run and the one you start by hand follow identical instructions. Avoid --yolo, which approves every tool call on its own, unless the run takes place inside a sandbox.

A monthly cron entry can be this simple, with the prompt text stored in a file you can edit without touching the schedule:

0 8 1 * * cd ~/linkedin && gemini -p "Run the monthly report prompt in prompts/11-report.md on the files in reports/data/" --output-format json > reports/latest.json

Run the command by hand first. If the output reports a missing file or a source it couldn't read, fix that before the schedule takes over.

Publishing stays manual by design: when a post from the prompt 6 calendar is due, you open LinkedIn and post it yourself.

You don't need 12 separate agents on the first day. Begin with the job that eats the most time and produces output you can judge reliably.

What can Gemini CLI do for your LinkedIn, and what stays with you?

Gemini CLI can organize your research, pull up useful evidence, prepare drafts and learn from your edits. Since everything lives in files and commands, the system is easy to inspect and run again. The quality still comes from the experience and judgment you put into it.

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

Related guides: how to build an AI content system for LinkedIn and LinkedIn maxxing with Claude Code.

Next step: create the LinkedIn folder, load your business context and run prompt 1. Better inputs make every later step more useful.