Appendix B: Skill Library

This appendix covers the skill pattern — how OpenClaw capabilities can be extended through packaged tool configurations, prompts, and integrations installed from ClawHub or defined directly in your context files. Skills are the bridge between raw tools and high-level behaviors, and understanding how to both use and build them unlocks a significant increase in automation sophistication. Rather than rebuilding common patterns from scratch every time, skills let you codify, version, and share specialized workflows.

The key distinction between skills and core tools is that skills are community-developed, independently versioned, and installable/removable rather than built into OpenClaw itself. They follow a trigger pattern — specific phrases or conditions that activate defined behaviors. The handbook demonstrates this through the Research Protocol (triggered by "research [topic]"), Data Pipeline Skill, and Browser Extraction Skill — all built from core tools without any installed package.

Building custom skills in context files (SOUL.md, AGENTS.md) is the most accessible entry point. Practitioners define triggers, inputs, process steps, outputs, and failure behaviors using a consistent template. The "Weekly Review Skill" example in the handbook shows how to wire together multiple data sources (task lists, memory logs, git activity, project directories) into a coherent automated workflow with explicit completion criteria.

The real power emerges when you combine the skill pattern with cron scheduling. Skills triggered by time-based schedules become persistent automation — morning briefings, nightly indexing, weekly reviews. ClawHub provides a registry of community-built skills for specialized domains, but the core skill patterns are universally applicable and require no external installation.

Key Items

• **Research Protocol** — Structured web research skill triggered by "research [topic]"; generates 5 search queries across different angles, fetches top 2 URLs each, synthesizes findings into structured reports saved to dated research files; template defined in SOUL.md for consistent execution
• **Data Pipeline Skill** — CSV/JSON transformation pattern triggered by data tasks; identifies input/output formats and locations, writes Python transformation scripts, validates output; foundation for ETL-style automations that process external data feeds
• **Browser Extraction Skill** — Triggered by "extract data from [URL]"; navigates with browser tool, snapshots the page, uses JavaScript evaluation to extract structured data as JSON, saves to appropriate data files; essential for sites requiring JavaScript rendering
• **Custom Skill Pattern** — Skills defined in context files by specifying trigger phrase, input sources, step-by-step process, output destinations, and failure behavior; the Weekly Review Skill example shows wiring together task lists + memory logs + git activity into a coherent workflow
• **Skill Routing in AGENTS.md** — Named agent routing rules that direct incoming requests to appropriate handlers based on keyword matching ("research *" → research-lead, "build *" → project-lead, "log *" → data-agent); enables multi-agent orchestration using skill triggers as routing signals

Related Concepts

• [[kelly-handbook-ch3-file-automation]] for file tools underlying skill patterns
• [[kelly-handbook-ch6-scheduling-and-cron]] for scheduling skills with cron triggers
• [[kelly-handbook-ch7-multi-agent]] for multi-agent coordination patterns
• [[karpathy-llm-wiki]] for the LLM wiki pattern this KB follows