Building Custom Agent Skills in OpenClaw

Out of the box, OpenClaw agents can read files, write code, execute commands, and call APIs. Custom skills extend these capabilities with domain-specific tooling and workflows.

What Are Skills?

Skills are documented capabilities that agents know how to use. They can be tool configurations (SSH access, database connections, API integrations), workflow patterns (how to deploy code, how to run tests), or domain knowledge (coding standards, design system documentation).

Skill Documentation

Skills are documented in workspace files that agents read. A TOOLS.md file might document SSH hosts, database connection strings, or API endpoints. The agent reads this documentation and uses the tools accordingly. Clear documentation means reliable tool usage.

Custom Integrations

Build custom integrations by providing agents with API credentials and documentation. An agent with Stripe API access and documentation can handle billing operations. An agent with GitHub access can manage pull requests. The integration boundary is whatever the agent can access from its workspace.

Shared vs. Agent-Specific Skills

Some skills are shared across the team (project coding standards, brand guidelines). Others are agent-specific (the developer agent knows how to deploy, the content agent knows the CMS). Organize skills accordingly — shared skills in project context docs, agent-specific skills in individual workspaces.

Extend your agents' capabilities: agentcenter.cloud