Kimi Code Skills: Getting Started with SKILL.md

Quick answer: Kimi Code is Moonshot AI's new terminal-based coding agent powered by K2.7 Code. It supports SKILL.md files for extensibility. Drop skills into your project directory and Kimi Code picks them up automatically.

Moonshot AI shipped Kimi Code in June 2026, adding another serious contender to the terminal AI coding agent space. Powered by the K2.7 Code model, it joins Claude Code, OpenCode, and Codex CLI as a CLI-first development tool. Here is how to set it up with SKILL.md skills.

What is Kimi Code?

Kimi Code (sometimes called KFC — Kimi for Code) is a terminal-based AI coding agent built by Moonshot AI. It uses the K2.7 Code model, which has shown strong performance on coding benchmarks, particularly for frontend development and structured code generation.

The agent runs in your terminal, reads your project context, and can write, edit, and debug code. It is built in TypeScript and designed for developer workflows that start from the command line.

Recommended skills

designing-hybrid-context-layers

by https://loreto.io · 16

Architects the right retrieval strategy for every query — teaching your agent wh…

$10

World-Class Site & App Design

by Aman Garg · 10

Every AI-built UI looks generic and templated. This skill teaches your agent to…

$5

ai-automation-qa-pack

by Roy Yuen · 10

Professional QA & UAT documentation generator for AI automation agencies and com…

$5

Browse all Kimi Code skills

Installing Kimi Code

Visit kimi.com/code and follow the installation instructions for your operating system. Kimi Code installs as a CLI tool and connects to Moonshot's API.

After installation, verify it works by running it in a project directory. It will scan your project structure and be ready to accept coding instructions.

Adding SKILL.md skills

Kimi Code supports SKILL.md files in your project's .skills/ directory, following the same convention as Claude Code, OpenCode, and other SKILL.md-compatible agents.

To install a skill from the Agensi marketplace:

1. Browse skills at agensi.io/skills and find one that fits your workflow
2. Download the skill package
3. Extract the SKILL.md file into your project's .skills/ directory
4. Kimi Code will detect the skill on its next run

For a quick start, try the Code Reviewer skill (free) — it teaches any compatible agent to run structured code reviews with severity-based findings.

What K2.7 Code is good at

K2.7 Code has emerged as a strong model for frontend-heavy projects. It handles React, Next.js, and Vue component generation well, and its structured output is clean. For backend and infrastructure tasks, Claude Opus 4.7 or GPT-5.5 generally produce better results, but K2.7 is competitive and improving.

The model is also available through OpenCode if you want to use it in a different agent environment. This gives you flexibility: use Kimi Code when you want Moonshot's native experience, or route K2.7 through OpenCode for its LSP integration.

Kimi Code vs Claude Code vs OpenCode

Kimi Code is newer and less battle-tested than Claude Code or OpenCode. Claude Code leads on code quality with Opus 4.7. OpenCode leads on flexibility with 75+ model support. Kimi Code's advantage is a fresh codebase and strong frontend performance with K2.7.

All three support SKILL.md, so your skill library works across all of them. For a full comparison of terminal agents, see the AI coding agent comparison.