Iteratively improve skill frontmatter compliance and test coverage using the Ralph loop pattern.

• Automates a 10-step feedback loop: read skill metadata, score compliance against the agentskills.io spec, scaffold missing tests, improve frontmatter triggers, run tests, validate references, check token budgets, and prompt for commit/issue creation
• Targets Medium-High compliance: distinctive WHEN: trigger phrases, descriptions under 60 words, passing tests, and token budgets under 500 lines
• Supports single skills, multiple skills, batch operations by adherence level (Low/Medium/Medium-High/High), and all skills at once; includes --skip-integration flag for faster iteration
• Validates skill names, description length, trigger keywords, and warns against risky "DO NOT USE FOR:" anti-triggers in multi-skill environments; preserves optional spec fields like license and metadata