kimi-agent-internals
Reverse-engineered Kimi agent prompts, skills, and runtime source
All Skills in This Repository (4)
Pure Emerald Level Indicators
dnnyngyenwebapp-building
Scaffold React web applications with TypeScript, Vite, Tailwind CSS, and shadcn/ui.
dnnyngyenxlsx
Create, style, and validate Excel workbooks with Python.
dnnyngyenConvert, extract, and edit PDF content using Python libraries.
dnnyngyendocx
Create and edit Word documents using OpenXML structures with Python-generated visuals.
Frequently Asked Questions
FAQPage SchemaHow to install kimi-agent-internals?▼
Run `npx skills add dnnyngyen/kimi-agent-internals --all -g -y` in your terminal to install all skills in this repository globally.
What is inside the Kimi agent internals repository?▼
It contains extracted system prompts for six Kimi agent types, 38 tool schemas, skill definitions for DOCX, XLSX, PDF, and webapp output, plus runtime source code and architecture analysis.
How does Kimi generate Excel and Word files?▼
The repository shows Kimi uses OpenXML-based skills with mandatory validation steps, formula error checking, and pivot table tooling to produce reliable spreadsheet and document files.
Is this repository official Moonshot AI code?▼
No. It is an independent research analysis of Kimi's production system, published under mixed CC0 and CC BY 4.0 licensing, and Kimi publicly confirmed the extracted code was genuine.
Who should read this Kimi architecture analysis?▼
Engineers building agent systems, prompt designers, and researchers who want to study how a production-grade agent handles skills, tools, and sandboxed execution.
Related Repositories in Education & Research
View All in Education & Research→
shareAI-lablearn-claude-code
Learn to build AI agent harnesses from scratch
microsoftai-agents-for-beginners
Learn to build, deploy, and secure AI agents step by step
leonardomso33-js-concepts
Master 33 essential JavaScript concepts with tested, verified guides