Social Media Automation with Claude Code
I built a system that handles my entire content workflow.
Before: 6 hours per video
Now: 45 minutes
Same quality. 10x more content. One simple dashboard.
Here's exactly how it works—and how you can set it up 👇🏻👇🏻
--- Author's self-reply in comments (full thread) ---
1/ What this system actually does
Most creators have the same problem.
Great ideas. Zero time to execute all of them.
You record one video. Then you spend hours writing the description. Designing the thumbnail. Writing LinkedIn posts. Cutting shorts. Uploading everywhere.
By the
2/ How it was built
I spent weeks building and refining individual Claude Code skills.
Think of a skill like a recipe.
One skill knows how to research YouTube topics.
One skill knows how to write scripts in my exact voice.
One skill knows how to generate thumbnails.
One skill
4/ Tab 2 — Video Creation (research + script + title in one place)
This is where every video starts.
Step 1 — Enter your idea
Type your topic. Select your target video length. Click create.
The system researches your idea automatically.
It shows you:
- What content already
4/ step 2 - Brainstorm with your AI
You can talk back to the research.
“Which SOP would be most universal and applicable to most businesses for a live demonstration?”
It thinks through your question and comes back with a recommendation with full reasoning.
You go back and forth
4/ step 3 - Generate your script
Tell it the direction you want. Click generate script.
It writes in YOUR voice because it was trained on your actual scripts.
It knows how you open videos. How you pace your delivery. What frameworks you use. What you never say.
You get hook
4/ step 4 - Generate title options
One click. Three title options.
Each one comes with:
- A rating on how likely it is to get clicked
- A thumbnail concept that matches the title
- The emotional hook it’s using and why it works.
5/ Tab 3 — Thumbnails (built on Gemini’s image model)
This tab uses Nano Banana — Google’s image generation model — because it produces better visual results than Claude for this job.
Option A — Remix a reference thumbnail
Upload a thumbnail you like.
Tell it what to change:
6/ Tab 4 — Publishing (the tab that saves an hour per video)
This is the most time-consuming part of uploading a video — and this tab eliminates almost all of it.
Here’s what used to take forever:
Writing the description. Formatting the timestamps. Adding tags. Writing the first
7/ Tab 5 — Repurposing (one video becomes content everywhere)
Two options for turning your long video into social content.
Option A — Paste the transcript
Copy and paste your transcript. Click generate.
Option B — Just paste the YouTube URL
The system transcribes the video
8/ The shorts pipeline the most impressive part of the whole system
This one runs from the command line as a standalone skill.
You give it a long video. It gives you finished vertical short-form clips — with captions — ready to upload.
Here’s the 6-step pipeline it runs
9/ The other tabs that complete the system
Calendar tab — Shows which videos are scheduled and what stage each one is at. You see your whole content pipeline at a glance.
Analytics tab — Pulls your YouTube data via API. Subscribers, total views, average views per video,
10/ What this actually cost to build
API keys needed:
- YouTube API — for research and analytics (free tier available)
- Google Gemini — for thumbnails via Nano Banana
- Deepgram — for video transcription
- Anthropic — for script generation and AI processing
- Blotato or