LinkedIn Post Generator Agent Template

What it does: Turns one idea into a finished, on-brand LinkedIn post (copy plus a generated image card or infographic) and saves it to Google Drive, with three approval checkpoints along the way.
Best for: founders and execs building a personal brand, marketing teams, social and content agencies, anyone posting on LinkedIn regularly
Apps used: Web Search, Fal AI image generation (Nano Banana Pro / Nano Banana 2), Google Drive
Setup time: 15 to 20 minutes (includes a one-time brand onboarding)

LinkedIn Post Generator is a no-code, six-agent workflow that turns one idea into a finished, on-brand LinkedIn post. An Orchestrator coordinates five specialists (research, copywriter, art director, editor, publisher), gathers proof with web search, writes in your saved voice, generates a branded image card, runs QA, and saves the post to Google Drive. This template acts as a starting point: any app, instructions, and agents can be updated as required for your own use case, simply ask Max (Agent Builder) to update the workflow as needed. Last verified from workflow config on June 2026. Includes common failure modes and fixes.

Problem this solves

Posting consistently on LinkedIn is slow: you research, write, design an image, and proof every post by hand.
Generic AI drafts sound off-brand and invent details, so you spend time rewriting and fact-checking.
Branded images usually need a designer or a separate template tool, and the text, logo, and colors rarely line up with the copy.
You want one place to go from idea to a finished, on-brand post, with control at each step instead of a black box.

What this agent does

This template can:

Qualify your idea and set one objective and one win condition before writing (Research agent).
Gather proof with web search and tag every item as user-provided, web-verified, or a flagged gap.
Write 8 to 12 hook options and a full draft in your saved voice (Copywriter agent).
Generate a complete on-brand image: a card or infographic with headline text, your logo, and brand colors in one render (Art Director agent).
Use your logo and a matching reference layout as image inputs so the structure matches your brand.
Hold a text-only veto when an image would not add value, with a written reason.
Run a voice and humanize pass, then QA the post against a fixed checklist (Editor agent).
Strip em dashes, en dashes, and banned phrases by literally sweeping the file before final review.
Assemble the final package (post text, first comment, hashtags, alt text, references) and save the image plus a text file to Google Drive (Publisher agent).
Keep you in control with three checkpoints: draft, image, and final save.

How it works

You onboard your brand once. Send setup and answer a short 4-turn interview: identity and audience, voice (paste 2 to 4 of your real posts), visual brand (upload your logo and 2 to 4 reference images), and CTAs and hashtags. The Orchestrator saves this profile and reuses it on every run.
You send one idea. Anything from one line to a full story, optionally with an objective and a target (personal profile or company page). Example: Why we killed our free trial | objective: authority | personal profile.
Research qualifies the idea, sets the objective and win condition, and builds verified proof with web search. It flags gaps instead of inventing facts.
Copywriter writes hook options and a full draft in your voice. Checkpoint A: you approve the draft or tell it what to change.
Art Director designs the visual concept and generates two on-brand image candidates (card or infographic) using your logo and colors. Checkpoint B: you pick one, request a change, or go text-only.
Editor runs the voice and humanize pass and the canonical QA, then assembles the final post. Checkpoint C: you review the finished post and reply save it.
Publisher uploads the image and one text file (copy, first comment, hashtags, references) to your Google Drive folder and writes a receipt. It does not post to LinkedIn.

Requirements

A connected image model: Fal AI with Nano Banana Pro (best) or Nano Banana 2 (renders the complete card or infographic, text and logo included, in one image).
A connected Google Drive account with upload access (a destination folder is optional; root is used otherwise).
Web Search: already enabled for the Research and Editor agents, so nothing to connect.
Brand inputs for onboarding: who posts, audience, voice, approved claims, 2 to 4 real LinkedIn posts to paste, brand hex colors, a logo file (transparent PNG or SVG), and 2 to 4 reference images.
An idea to start each run. The more real material you give (numbers, stories, links), the fewer gaps the pipeline has to flag.

Setup guide

Clone the template and open it in AgentSpace.
Connect the Fal AI image node: click the node, connect, and confirm Nano Banana Pro or Nano Banana 2 is available (the template allows both).
Connect Google Drive: click the Google Drive node, connect, and authorize file upload. Note or pick the destination folder.
Web Search is already enabled for the Research and Editor agents; nothing to connect.
Run onboarding: send setup in the Playground and answer the 4-turn interview (identity, voice with 2 to 4 pasted posts, visual brand with logo and reference uploads, CTAs and hashtags). Approve the profile when prompted.
Write your first post: send a one-line idea, for example Why we killed our free trial | objective: authority | personal profile.
Move through the three checkpoints: approve the draft, pick an image, then reply save it at final review.
Confirm the image and the text file land in your Google Drive folder.

Common issues and fixes

Image node not connected: the Art Director fails fast with "no image tools connected." Connect Fal AI (Nano Banana Pro or Nano Banana 2) and re-run from the art-director step.
Logo or reference not used in the image: during onboarding, upload the logo file and 2 to 4 reference images so they become logo_url and reference_image_urls in the profile. A LinkedIn post link is not a usable reference; only an uploaded image file works.
Misspelled text or wrong numbers in the image: the Art Director inspects each render letter by letter and regenerates, up to a budget of 6 generations. If in-image text still fails, accept the text-only option or shorten the on-image copy (6 words maximum).
"No LinkedIn profile" failure: every agent gates on the saved profile. Run setup first to create the profile before you request a post.
Post not saved to Drive: the Publisher only runs after you reply save it and the approval line is logged. If Google Drive is not connected, the Publisher stages the full package in the receipt for a manual save instead of failing silently.

Customization knobs

Swap the image model (Nano Banana Pro vs Nano Banana 2) in the Fal AI node.
Change agent models: the Orchestrator and Art Director run Claude Sonnet 4.6; Research, Copywriter, Editor, and Publisher run DeepSeek V4 Flash. Adjust per cost and quality.
Keep brand facts (colors, voice, approved claims, hashtags) in the profile, not in the prompts. Use update profile to change them.
Turn on live LinkedIn posting later by adding upload-post target details during onboarding (off by default; Drive delivery is the default).
Send fully unattended in the brief to skip the draft and image checkpoints. The final save checkpoint always stays.

Apps used

App	What it is used for	Typical permission scope
Web Search	research and proof-gathering for the post (steps 01-02) and claim re-verification (step 09)	internet search (built in)
Fal AI image generation (Nano Banana Pro / Nano Banana 2)	render the complete on-brand post image (card or infographic with headline, logo, and motif) in one shot, using your logo and a reference layout as inputs	image generation API (list_image_models, generate_image)
Google Drive	upload the finished image and a text file (copy, first comment, hashtags, references) to your Drive folder	file upload, create folder and file (OAuth)

(Scopes vary by how you connect each app. The image and Drive nodes are required; web search ships enabled.)

Use cases

Founder thought-leadership: Turn a point of view into an authority post with an on-brand image card. Best for founders and execs building a personal brand.
Company page content: Produce on-brand posts for a company page with consistent voice and visuals. Best for marketing teams.
Product and launch announcements: Pair a clear message with a branded image so launches look consistent. Best for product and growth teams.
Agency content production: Onboard each client brand once, then produce posts that match their voice and visuals. Best for social and content agencies.
Repurposing wins and data points: Turn a verified stat or result into a claim card or infographic. Best for marketers who post regularly.

Example prompts and outputs

Starter sequence (copy/paste)

First run: setup (answer the short interview, then approve your profile).

Then send one idea per run, for example:

Why we killed our free trial | objective: authority | personal profile

Example 1: Idea to finished post

Prompt (input): Why we killed our free trial | objective: authority | personal profile

Output: Research builds proof and flags gaps, Copywriter presents a draft with hook options at Checkpoint A, Art Director presents two on-brand image candidates at Checkpoint B, and after you reply save it the final post (text, first comment, hashtags, and image) is saved to your Google Drive.

Example 2: Fully unattended

Prompt (input): 3 lessons from migrating to event-driven architecture | fully unattended | company page

Output: the pipeline skips the draft and image checkpoints, builds the post and image automatically, then stops at final review and waits for save it. The save step is never skipped.

Example 3: Data post with real numbers

Prompt (input): Our support reply time dropped from 9 hours to 40 minutes after we added an AI receptionist | objective: authority

Output: an infographic-style card that shows the exact before and after numbers (used only because you provided them in the brief), with copy and hashtags, saved to Drive. The pipeline never invents numbers; unsupported figures are flagged as gaps instead.

Why you need this template

One workflow from idea to finished post: copy plus an on-brand image, saved to Drive in one run.
Consistent brand voice and visuals from a saved profile, reused every run, instead of re-explaining your brand each time.
Three human checkpoints keep you in control of the message, the image, and the save.
No fabrication: facts, numbers, and claims come only from your approved claims, your brief, or web-verified sources, with gaps flagged honestly.
Editable in AgentSpace: change models, swap the image model, or update brand facts without code.