skill-developer

Creating a Claude Skill involves defining its file structure, writing semantic metadata, configuring triggers for automatic activation, and testing against the 500-line rule for token efficiency. Start by establishing the skill's core task, keywords, and dependencies, then structure content for progressive disclosure.

Triggers are the mechanisms that automatically activate your Skill based on user queries. Configure triggers by defining precise skill descriptions and keywords that match search intent. The two-tier injection model routes meta-skills and domain-skills through distinct trigger pathways for optimal relevance.

The 500-line rule enforces token efficiency by requiring progressive disclosure: expose core functionality upfront, then progressively reveal advanced details. Structure content into sections, use references for external detail, and test retrieval efficiency to ensure your Skill activates precisely without exceeding token budgets.

Meta-skills provide foundational automation patterns and governance—like this Skill itself—while domain-skills encapsulate specialized knowledge in specific technical areas. The two-tier model ensures meta-skills inject contextual rules first, then domain-skills activate based on user intent and skill metadata.

Maximize discoverability by crafting clear, searchable keywords that describe your Skill's core task, not its structure. Write a concise tagline, use semantic metadata with vector-embedding-friendly fields like request and description, and configure trigger activation to match genuine user search vocabulary.

Session tracking enables hook mechanisms that maintain context across multiple Skill activations, allowing progressive disclosure and stateful workflows. Document session scope in your Skill metadata to ensure the two-tier injection model correctly sequences meta-skill rules and domain-skill behavior.