Claude's entire system prompt leaked to GitHub.
Clone it into a free model and run your whole agency for pennies.
Take the 3,800-line rulebook that makes Claude reason and paste it into a free open-source model so it inherits the same discipline at near-zero cost per token.
What's inside
- The exact repo link with the full leaked prompt file
- Step-by-step setup for Kimi K2 or GLM
- Custom-agent config to reuse across every client
- Three workflows: client work, outreach, and content
The Free Frontier Clone: Run a Near-Frontier Assistant at Zero Token Cost
by Divjot Sahni · EZYE Consulting · ezye.com.au
What actually makes a frontier model 'smart'
Most solopreneurs pay for the frontier model when the real magic is the instructions running underneath it
The system prompt controls reasoning style, tool-calling discipline, and citation habits
Here's the part almost nobody realizes about why cheaper models feel 'dumber'...
Copy the right rulebook and the gap between a $20/mo model and a free one nearly disappears
Dropping the leaked prompt into a free model
Grab the file from the repo (linked after you unlock)
Open a free open-source model like Kimi K2 or GLM
Paste the leaked prompt in as your system message
Then save it as a reusable custom agent so every chat inherits it...
Free access — enter your email below