April 23, 2026 / Self-developed Tools / Read Time: 6 Min

Possibly the First Free Chinese Law Verification SKILL — Use It Freely

Introduces a free Chinese law verification SKILL that lets AI agents check the accuracy of legal provisions during reasoning, preventing AI hallucinations, with installation instructions and a SKILLHub link.

AI-generated laws and regulations can hallucinate.

This has been a persistent problem in the legal industry.

Especially as agents develop rapidly—

If the laws are wrong midway, many downstream conclusions will also be wrong.

Let alone building stable workflows.

So I built a SKILL myself.

And set up my own legal database.

It contains correct, current laws and regulations.

Legal companies charging 0.5 to 1 yuan per API call?

I’m giving it away for free.

(Until I can’t afford it anymore.)

  • This article represents only the author’s personal views and should not be considered legal advice.

I. See It in Action

Verify the authenticity of legal provisions

In Hermes Agent, ask AI to verify whether relevant laws are correct:

Search legal provisions

Query legal provisions directly (my public-interest API doesn’t support fuzzy search yet):

Prevent hallucinations

II. How to Install?

We’re in the agent era, so installation is very simple. Just tell your AI:

Visit https://skillhub.lslby.com/llms.txt and install the chinese-law-verifier skill

And it’s installed.

III. What Else?

In the Agent era, you can’t be without your own SKILLHub. So I built one:

https://skillhub.lslby.com/

I’ll occasionally update original SKILLs there, and I promise:

They certainly won’t be the kind that wastes your AI agent’s tokens watching ad-filled WeChat articles.

Feel free to have your agent periodically check:

https://skillhub.lslby.com/llms.txt

For any useful new SKILLs.

Besides this verification SKILL, there’s also one that converts long screenshots into paginated PDFs.

It’s very useful when you’ve collected a bunch of extremely long screenshots—like web page screenshots, WeChat chat history screenshots—that need to be converted to PDFs as evidence.

Essentially, it’s the SKILL version (long screenshot only) of:

Still Manually Splitting Chat Records for Evidence? Time to Automate | Automated Evidence Screenshot and PDF Generation Based on PaddleOCR

See the article above for detailed results.

IV. Final Thoughts

Features like this should really be done by the authorities.

After all, they have the most complete legal database.

But even if they did, it would probably get DDoS’d to death by legal companies.

Same as always: if the authorities won’t do it, I will.

The server is still being optimized. I’ll release more legal search capabilities later depending on how things go.

For now, let’s see if any legal company wants to DDoS me.

Hope my optimization can hold up.

Boyang Li
Author

Boyang Li

Chinese Attorney — Beijing Longan (Guangzhou) Law Firm

A lawyer focused on game law, AI regulation, data compliance, and digital content rights. I write about practical legal insights for innovative tech teams.

Contact me about this topic →

I Trained a Better Legal Embedding Model Than Google and Alibaba?

Sharing the complete process of fine-tuning a legal-specific embedding model: based on Google's EmbeddingGemma-300M, trained on a legal provision-to-colloquial question dataset, outperforming Google and Alibaba's models in legal provision retrieval, now open-sourced.