n8n workflow: new blog post → channel-adapted posts across LinkedIn, Twitter, Instagram, Facebook, with a review queue.
A content team publishes a blog post and then spends 60-90 minutes rewriting it for four social channels — different tones, different lengths, different image sizes. The cross-posting is mechanical but the channel-voice adaptation isn't, and copy-paste accidents (LinkedIn tone on Twitter) are common.
n8n workflow triggered by RSS or webhook from the CMS. Claude rewrites the post per channel — LinkedIn long-form, Twitter thread, Instagram caption, Facebook short — anchored by a brand-voice system prompt + few-shot exemplars per channel. Hero image is reused with per-channel resize where needed. Variants land in a review queue; approved items publish, others sit until a human edits or drops them.
Per-post adaptation drops from ~75 min to ~10 min of review. Brand voice stays consistent because the few-shot exemplars are checked into the workflow, not in someone's head.
Telling the model 'be conversational but professional' is fragile. Showing it five real prior posts that hit the voice is robust — and when the voice drifts, you update exemplars rather than wordsmithing the prompt.
At this team's volume (~2 posts/week), the marginal cost of human review is small and the cost of a bad auto-post is large. The workflow optimizes for catch-rate, not throughput.
Each social platform is a swappable n8n node. Adding TikTok or Threads is a new node + a new prompt — no changes to the orchestration.
Run the free 5-minute diagnostic — I'll tell you whether it's a $0 weekend project or a 2-week build.