An advanced AI coding executor that enables agents to generate, refactor, and debug complex multi-file projects using the Kimi Code CLI.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install kimi-cli
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install kimi-cli using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Kimi CLI Coding Agent Skill?
The Kimi CLI skill transforms an AI agent into a sophisticated coding architect. Instead of just writing code snippets in a chat, this skill allows the agent to command the Kimi Code CLI to produce or modify actual files within a local environment. It leverages the OpenClaw exec PTY mode to handle interactive terminal sessions, ensuring stable output and the ability to perform long-running development tasks such as project scaffolding and legacy code modernization.
By integrating this skill, users of Openclaw Skills can offload high-complexity implementation details to a specialized execution engine while the main session model focuses on planning, requirement clarification, and final acceptance testing. It is designed for scenarios where task complexity exceeds the limits of a single model response.
Kimi CLI Coding Agent Skill Use Cases
- Large-scale code generation involving full functional modules or multi-file project structures.
- Complex code refactoring including modernization of legacy code and architectural adjustments.
- Deep bug diagnosis requiring analysis of errors across multiple interrelated files.
- Technical research and PoC development for integrating new technology stacks.
- Long-running development tasks that require continuous interaction and persistent session state.
How Kimi CLI Coding Agent Skill Works
- The session model (e.g., GLM) plans and breaks down the technical requirements into actionable steps.
- The agent invokes the Kimi CLI skill using either Quick Mode for one-shot tasks or Interactive Mode for multi-step processes.
- The skill utilizes OpenClaw exec with PTY (Pseudo-Terminal) enabled to initialize a terminal session in a designated working directory.
- Kimi Code CLI executes the specialized command, generating or modifying files and providing step-by-step execution logs.
- For interactive tasks, the agent uses process control actions like submit, log, and poll to manage the background lifecycle until the task is complete.
- Upon completion, the agent verifies the output using automated checks or Ralph-style summary reports.
Kimi CLI Coding Agent Skill Setup
To get started with this skill in Openclaw Skills, follow these installation steps:
# Install the Kimi CLI via pip
pip install kimi-cli
# Authenticate your account
kimi /login
# Ensure your target directory exists before calling the skill
mkdir -p ~/my-project-dir
# Example: Running a one-shot project generation task
bash pty:true workdir:~/my-project-dir timeout:300 command:"kimi --print -p 'Create a React + TypeScript Todo application'"
Kimi CLI Coding Agent Skill Data Schema & Taxonomy
The skill organizes its output and metadata to ensure compatibility with various Openclaw Skills workflows:
| Component | Description |
|---|---|
| Workdir | The isolated directory where Kimi CLI creates and modifies project files. |
| Session ID | A unique identifier (e.g., abc123) used to manage background interactive processes. |
| payload.acp | An ACP-compliant output file containing versioning, signatures, and task data for validation. |
| summary.json | A machine-readable JSON file detailing metrics and execution status (e.g., SUCCESS). |
| Session Logs | Real-time terminal output captured via the PTY session for debugging and monitoring. |
Kimi CLI Coding Agent Skill Advanced Features
- Full PTY (Pseudo-Terminal) support to ensure interactive CLI tools behave correctly within Openclaw Skills.
- Interactive Mode for background execution, allowing agents to respond to prompts (e.g., confirming file overwrites).
- Robust process management via actions including log, submit, write, poll, and kill for fine-grained control.
- Standardized 429 Rate Limit Handling SOP to manage API quotas effectively without failing tasks.
- Ralph-style output integration, generating consistent summary.json and summary.txt files for downstream agent parsing.
- Multi-language support with best practices for python3 and shell environments to prevent dependency conflicts.
SKILL.md
Loading
Related Openclaw Skills
Ralph: AI CLI Automation Loop Generator
Ralph is a specialized command generator that creates robust automation loops for running AI CLI tools like Claude Code and Gemini through local terminal environments.
Crypto Whale Monitor
A specialized blockchain monitoring skill that uses Web3 RPC to track large-scale liquidity movements from whale wallets.
multisage
A powerful multi-model synthesis tool that queries Claude, GPT, and Gemini simultaneously to provide comprehensive, expert-vetted answers.
Cursor CLI Passthrough (cursor-tui)
A transparent bridge that spawns and relays Cursor Agent's CLI sessions directly to the user, making it one of the most essential Openclaw Skills for terminal-based AI development.
Trust Guard
A security enforcement layer that scans and scores every skill before installation to prevent malicious code execution.
YouTube Lecture Transcript Analyzer
A powerful CLI tool to transform YouTube video transcripts into structured lecture notes, evidence-backed claims, and actionable checklists.
Featured*
