skill-developer

To create a custom AI Skill, start by making a directory for your Skill and adding a SKILL.md file with YAML frontmatter containing `name` and `description` fields. Write your operational instructions in the Markdown body, then optionally add `scripts/`, `references/`, or `assets/` directories as needed to structure your Skill according to platform standards.

Define Skill metadata including type, enforcement rules, priority, triggers (prompt and file-based), skipConditions, blockMessage, and versioning. The skill-rules.json configuration file captures these requirements and validates your Skill against the complete schema to ensure it meets platform specifications.

Triggers define when a Skill activates—either by matching prompts or monitoring files. Hooks enable dynamic Skill behavior and integration by intercepting execution at key points. Together, they allow you to automate workflow responses and enforce rules consistently across domain and guardrail contexts.

Common Skill development issues include schema validation failures, trigger misconfigurations, and enforcement rule conflicts. Use the troubleshooting guide to diagnose problems by checking SKILL.md structure, validating skill-rules.json against the schema, and testing multi-trigger scenarios systematically.

Yes. The platform supports multi-trigger testing to validate that prompt and file triggers work as intended. Test your trigger patterns thoroughly before deployment to ensure they activate correctly and enforce your intended rules in all expected contexts.

A patterns library provides proven design approaches for recurring Skill challenges. These patterns guide you through structuring progressive-disclosure documentation, managing guardrail contexts, and handling complex trigger logic so your Skills remain maintainable and platform-compliant.