Claude Code Skills on ClawHub: The Complete Directory for Orchestration, Automation & More
There are dedicated Claude Code skills on ClawHub — and they're not what you might expect.
Unlike typical platform integrations, these skills don't add features to Claude Code. They work the other way around: they let you control Claude Code from OpenClaw, dispatching it as an autonomous coding agent from any connected chat interface — Telegram, QQ, WeChat, Discord, or a terminal. The entire category is built around one insight: Claude Code is a powerful local agent, and OpenClaw is the command layer on top.
Note: Install and download figures in text descriptions reflect stats at the time of writing and may be outdated. All skill tables are live — they fetch current data from the ClawHub database on every page load. Treat table values as authoritative.
Data Overview
| Metric | Value |
|---|---|
| Dedicated Claude Code skills | 58+ |
| Skills with installs > 0 | ~25 |
| Top by downloads | claude-code-wingman ( downloads) |
| Top by installs | claude-code ( installs) |
| Zero installs | ~33 |
1. Orchestration & Multi-agent
The biggest and most popular category. These skills spin up Claude Code instances inside tmux sessions, coordinate parallel workers, and manage task queues — turning one Claude Code installation into an always-on coding engine that OpenClaw can dispatch work to asynchronously.
claude-code-supervisor (installs: ) is the cornerstone: it watches Claude Code sessions running in tmux and relays results back. claude-code-wingman (downloads: ) takes this further by orchestrating multiple Claude Code instances in parallel — one directing, others executing. claude-tmux is the lower-level primitive that most orchestration skills build on.
2. OpenClaw ↔ Claude Code Bridge
The core integration layer — skills that establish the connection between OpenClaw's messaging runtime and the Claude Code CLI on your machine. Some wrap the CLI directly as a subprocess; others use ACP (Agent Communication Protocol) or MCP as the bridge layer.
claude-code and openclaw-claude-code are the two flagship bridge skills, both with installs, handling the basic "send a task to Claude Code and get a result back" workflow. cc-bridge and claude-code-bridge extend this to messaging platforms like QQ, Telegram, and WeChat. discord-claude-code-delegation does the same for Discord.
3. Usage & Token Management
Quota anxiety is real for Claude Max subscribers. This category exists entirely because of it. These skills monitor session usage, warn before limits are hit, refresh expired OAuth tokens automatically, and find ways to squeeze more out of each token budget.
claude-code-usage (downloads: ) is the most downloaded skill in this category — it checks OAuth session limits and weekly quotas in real time. claude-code-pro takes a different approach: instead of monitoring, it just burns fewer tokens by optimizing prompts before they reach the model.
4. Setup, Config & Model Switching
Getting Claude Code production-ready is non-trivial. This category covers environment setup, CLAUDE.md initialization, settings management, and — notably — skills for routing Claude Code tasks through alternative models (MiniMax, Bailian, Codex) when Anthropic's API is too expensive or rate-limited.
claude-code-setup scaffolds a .claude/ collaboration layer for any new project. claude-code-minimax and openclaw-claude-code-bailian are the model-switching options — useful if you're in China or need a cheaper backend.
5. Security & Skill Auditing
A small but important category. As the Claude Code skill ecosystem grows, so does the risk of malicious or poorly-written skills. These tools let you audit skills before installing them, check shell commands for injection risks, and get recommendations for which Claude Code automations are worth setting up.
6. Specialized Task Extensions
Skills that extend Claude Code with a specific domain capability — web search, browser automation, Twitter/X scraping, batch codebase migrations, and Anthropic API building. These are closer to traditional integrations: Claude Code handles the coding logic, and the skill provides the domain-specific tooling.
7. Learning & Best Practices
Reference skills that don't run code — they guide you on how to use Claude Code more effectively, configure MCP tool servers, and optimize your workflow architecture.
Quick Reference: What Do You Actually Need?
| I want to… | Recommended skill |
|---|---|
| Send coding tasks to Claude Code from Telegram/QQ | claude-code or openclaw-claude-code |
| Run multiple Claude Code agents in parallel | claude-code-wingman |
| Watch a Claude Code session in tmux | claude-code-supervisor |
| Check my Claude quota before it cuts out | claude-code-usage |
| Set up a new project for Claude Code | claude-code-setup |
| Use a Chinese model instead of Anthropic | claude-code-minimax or openclaw-claude-code-bailian |
| Audit a skill before installing it | skills-security-check |
| Add web search to Claude Code | claudecode-websearch |
A Few Closing Observations
These aren't Claude Code plugins — they're the inverse. Every skill in this directory treats Claude Code as the worker, not the platform. OpenClaw is the command layer; Claude Code is the execution engine. That's a fundamentally different pattern from, say, a Notion integration skill.
Usage monitoring is disproportionately popular. Three of the top ten skills by download are about quota tracking. Claude Max's session limits are a genuine pain point, and the community has responded with a whole sub-ecosystem of workarounds.
The zero-install rate is high (~57%), but not surprising. Most of these skills require a local Claude Code installation with specific OS configurations (tmux, macOS Terminal access, etc.). They're not casual installs — they're infrastructure. A user who hasn't set up their environment won't install them.
The multi-model skills signal something real. claude-code-minimax and openclaw-claude-code-bailian exist because Anthropic's API is either blocked or too expensive for a significant portion of the user base. The demand for "Claude Code behavior, non-Anthropic model" is real and growing.
Data source: ClawHub platform install and download stats as of April 4, 2026. Browse more skills at clawhub-skills.com.