Startup growth agent

LinkedIn image posts

Renders a branded image for the day’s post and attaches it automatically.

Running since Oct 6, 2026·LinkedIn·Risk: Medium

What this solves

A text-only post is easy to scroll past. A visual with no extra editorial effort is the cheapest lift available for engagement, but it is also the first thing to get skipped when a posting workflow is already running unattended.

Who it’s for: Anyone already running the LinkedIn posting agent who wants every post to carry an image without adding a design step.

How we use it: Shipped today, as an extension of the LinkedIn posting agent rather than a separate thing to run: every post it publishes now goes out with a branded image attached, generated from the same draft in the same run.

How it works

What happens on every run.

  1. Takes the day’s post draft — the same text the LinkedIn posting agent just wrote — and reads its first line as the headline.
  2. Renders a 1080×1080 branded card in code: the company wordmark, the headline, and the measure, execute, remeasure loop drawn as three connected boxes. No external image API, no network call.
  3. Hands the image to the LinkedIn posting agent’s browser automation, which attaches it to the composer right after typing the post text.
  4. Clicks through LinkedIn’s media review screen (crop and alt text, with its own Back and Next buttons) before the normal Post button appears.
  5. If image generation or the attach step fails for any reason, the post still goes out text only rather than being blocked on a picture.

Skills and scripts

What it’s built from.

A Node script that renders the card (the same offline, no-external-API tool — @vercel/og — the site’s own social-share images use) plus an added step in the LinkedIn posting agent’s browser automation. Not packaged for public download yet — the setup guide below describes what it does; the managed tier runs it for you today.

Setup guide

Running it yourself.

Runtime

  • Everything the LinkedIn posting agent already needs — this runs as an extra step inside that same workflow, not as a separate agent to invoke.
  • Node.js, for the image-rendering script (no browser, no network call — it writes a PNG to disk).

Browser setup

  • No separate browser setup. The image is attached inside the same Chrome session and the same composer the LinkedIn posting agent already drives.

Account setup

  • No accounts beyond the LinkedIn posting agent’s own company page access.

Configuration

  • A fixed branded template today — wordmark, accent color, and the measure, execute, remeasure loop — not something configured per post.
  • The headline comes straight from the post’s first line; there is no separate image brief to write.

First run

  • Generate one image on its own and look at the PNG before wiring it into a real publish.
  • Run one real post with it and confirm LinkedIn’s media review screen gets clicked through automatically rather than stalling before Post.

Common questions

LinkedIn shows an image review screen with Back and Next buttons instead of Post. What happened?

That is expected — the agent clicks through it automatically. If a future LinkedIn redesign changes that screen, it is the one selector in this step most likely to need updating.

The image failed to attach. Did the post still go out?

Yes, text only. The attach step is deliberately best effort: a broken selector costs the picture, never the post.

Risk notice

What to watch for.

  • Shares its risk profile with the LinkedIn posting agent — the image step does not meaningfully add to LinkedIn’s own automation detection.
  • LinkedIn’s media review screen is one more place a redesign can break the flow; we default to publishing text only rather than blocking on it.

Run it

Do it yourself, or have us run it.

Run log for this agent: coming soon