What problem does it solve?
This Skill solves the common pain points of building AI skills that waste tokens through unnecessary over-loading, fail to trigger correctly for relevant user tasks, break across different AI agents like Claude Code and GPTel, or become unmaintainable monolithic knowledge dumps. It provides a complete, validated methodology for engineering skills that adhere to the Agent Skills Specification and Two-Agent Model for optimal performance and portability.
Core Features & Use Cases
- End-to-End Skill Creation SOP: A mandatory 5-phase process for building new skills from ground truth mining to testing, ensuring accurate routing and complete rule coverage.
- Skill Refinement Patterns: Proven workflows for splitting monolithic skills into atomic rules, updating existing skills without breaking routing, and promoting stable session guidelines into permanent skill rules.
- Multi-Agent Compatibility: Guidance for setting up skills to work seamlessly across Claude Code, GPTel, Cursor, and other agents, including symlink configuration and context budget management.
- Rule Authoring Best Practices: Standards for writing atomic, self-contained, exhaustive rule files with correct YAML frontmatter for accurate analyzer routing.
- Use Case: If your team has a 1200-line monolithic skill that loads all its content on every request, use this Skill to split it into 6 focused rule files that load only when needed, reducing token waste by over 60% while improving output accuracy.
Quick Start
Use the skill-architect skill to create a new skill for your team's API design conventions, following the 5-phase creation SOP to ensure it routes correctly and loads only the relevant rules for API-related tasks.
