Getting Started
Learn the basics of OpenClaw Skill and how to install your first skill bundle. Understand the ecosystem and discover what's possible with custom automation.
Read Guide →Create custom skills tailored to your workflow with simple SKILL.md files. No fine-tuning needed—just natural language instructions that teach AI agents exactly what you need. Open source, community-driven, zero barriers.
$ npx openclawskill install soul-personality
✓ Installing skill: soul-personality@2.4.1
✓ Dependencies resolved
✓ Skill installed successfully
Versioned, rollback-ready, discoverable through intelligent search
Install any skill folder in one command:
npx openclawskill install soul-personality
Curated excellence — highlighted for immediate trust
Core personality and values system. Defines agent behavior with explicit instructions: "be genuinely helpful, have opinions, be resourceful." No fine-tuning needed—just well-crafted prompts.
File-based memory architecture. Reads memory/YYYY-MM-DD.md for recent context and MEMORY.md for long-term memories. No vector database—just markdown files that update as the agent learns.
First-run experience that creates the "coming alive" moment. Guides agents through identity discovery with "Who am I? Who are you?" prompts. Theatrical but effective for memorable first impressions.
Dynamic identity system where agents define their own name, creature type, vibe, and emoji during first conversation. Creates unique personalities through structured markdown templates.
Minimal Telegram bot adapter connecting messaging platforms to Claude Code. Demonstrates core OpenClaw messaging patterns: receive message, call Claude, send response. ~100 lines of Python.
Deep dive into system prompt construction. Covers base identity injection, tool definitions, skill discovery, memory recall, and heartbeat systems. Essential for understanding OpenClaw behavior.
Fresh uploads across the entire registry
Environment-specific tool configuration system. Manages camera names, SSH hosts, speaker names, and other context-aware device mappings for seamless integration.
User information management system. Stores name, timezone, preferences, and contextual details to personalize agent interactions and maintain conversation continuity.
Workspace behavior rules and safety guidelines. Defines when to speak in group chats, memory system protocols, and interaction boundaries for Claude Code integration.
Intelligent group chat behavior system. Implements HEARTBEAT_OK protocol, reaction vs reply logic, and conversation dominance prevention through explicit prompt rules.
Proactive check-in and heartbeat mechanism. Enables agents to initiate conversations, check on users, and maintain engagement without being intrusive or dominating chats.
Platform-specific markdown formatting adapter. Converts markdown to platform-native formats for Telegram, WhatsApp, Discord, Signal, and other messaging channels.
Natural language instructions, not rigid APIs
---
metadata.openclaw:
emoji: 🍷
requires:
bins: [sqlite3]
env: [CELLAR_DB_PATH]
---
# Wine Cellar Manager
Manage your wine collection with ease.
Track vintages, regions, tasting notes...
## Usage
When the user asks about wine...
Confirm additions, validate data...
## Implementation
Uses SQLite database at...
Syncs with external services...
Every skill lives in a folder with one essential file: SKILL.md. This Markdown file teaches the AI agent how to use your skill through natural language instructions rather than rigid API documentation. Start with the skill name, add a description, include usage examples, and provide implementation details.
The metadata.openclaw YAML block configures how OpenClaw loads and manages your skill. Specify emoji icons, dependencies (bins, env, config), and installation commands. This ensures your skill works reliably across different environments.
Write instructions in plain English describing behavior, edge cases, and preferences. The AI reads these instructions and adapts its behavior accordingly. No rigid schemas. No extensive boilerplate. Just clear communication about what the skill should accomplish.
Share your AI capabilities with the community
Every skill lives in a folder with one essential file: SKILL.md. Write natural language instructions explaining what your skill does, when to use it, and how it works. No rigid schemas—just clear communication.
Configure the metadata.openclaw block with emoji icons, dependencies (bins, env, config), and installation commands. This ensures your skill works reliably across different environments.
Test your skill thoroughly with various prompts. Verify dependency installation on clean systems. Document configuration clearly, then share with the community.
Build exactly what you need, expressed in your terms
Create skills for workflows that generic tools can't handle. Wine cellar inventory, PR review automation, cross-platform publishing—specialized workflows that match your exact needs.
Every unique workflow you automate compounds your productivity advantage. Your personal OpenClaw instance becomes uniquely adapted to your work patterns over time.
Enable agent workflows that span extended timeframes. Monitor conditions, take actions, and report results over days or weeks—impossible with session-based tools.
No deep machine learning knowledge required. Just clear thinking about your workflow and the ability to express it in natural language instructions. Start small, iterate, and share.
Learn the basics of OpenClaw Skill and how to install your first skill bundle. Understand the ecosystem and discover what's possible with custom automation.
Read Guide →Master the anatomy of SKILL.md files. Learn how to write natural language instructions that teach AI agents to use your skill effectively through plain English descriptions.
Read Guide →Configure the metadata.openclaw block with emoji icons, dependencies, environment variables, and installation commands. Ensure cross-environment compatibility.
Read Guide →Learn when to build custom skills versus using existing ones. Understand domain-specific needs, specialized behaviors, and integration depth requirements.
Read Guide →Study successful community skills like wine cellar management and PR review automation. Learn patterns for robust implementations and graceful edge case handling.
Read Guide →Best practices for testing skills in isolation, verifying dependency installation, documenting configuration, and sharing with the community following guidelines.
Read Guide →