Anatomy of a Well-Designed OpenClaw Skill: Frontmatter, Triggers, and Compatibility

Short summary

This guide breaks down how to write OpenClaw agent skills that agents actually match and load correctly, focusing on frontmatter design, trigger conditions, and compatibility gating. Using real pilot-* skills as worked examples, it shows how specific descriptions with concrete verbs, nouns, and literal CLI strings help agents pick the right skill among near-duplicates, while negative trigger conditions prevent unnecessary loads. It also covers compatibility fields like allowed-tools, runtime notes, and metadata.openclaw.requires that gate eligibility based on the host environment.

• •Write matchable descriptions with concrete verbs, nouns, and literal strings — not vague taglines
• •Use disambiguating language so agents can tell similar skills apart and rule out the wrong ones
• •Gate eligibility with allowed-tools, compatibility notes, and metadata.openclaw.requires to prevent misfires on incompatible hosts

Generated with AI, which can make mistakes.