Schedule and publish verified AI news posts

An automated agent checked recent AI stories, created image posts, and pushed them to Xiaohongshu after a manual login when the session expired.

Done with Custom agent· Advanced· Oct 2026

What they did
A cron job ran at 09:00 and woke the agent. It searched the web several times for AI news from the last day or two, opened the original pages to confirm the details, and assembled a spec file with five points and their URLs. Scripts then converted that spec into 1080×1440 images, and a browser workflow using a pre-authenticated Edge profile plus Chrome DevTools Protocol handled uploading, filling in the post, and publishing. Afterward, it verified that the post had gone live.
How it went
During 36 days of operation, 28 out of 29 scheduled runs published successfully, with no duplicate posts. The one miss came from a model API timeout. Each attempt finished in about 4 to 8 minutes, and the hashtags were posted as ordinary text rather than clickable links.
Tips
Publishing was fragile: the publish control lived inside a closed shadow DOM, mouse-coordinate clicking could fail unpredictably, and the editor did not respond to keypresses. The repository lists 20 separate failure modes, so a fallback path is important.

Try it yourself with OpenClaw

At [time] each day, look for the [number] most relevant recent stories about [topic] from the last [24–48] hours. Confirm each story against its source page and one more reliable source, then draft a short [language/tone] post with 5 bullet points and the URLs. Render the post as a [size] image card, then stop and ask for approval before you do anything that would send, publish, delete, or pay for something. If approved, publish it to [platform], confirm it is visible, and record the final text, the links, and any errors.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.