Improve skill quality scoring algorithm

[NeuZhou]

Good First Issue

The quality scoring algorithm could be more accurate and useful.

Current Issues

• Equal weighting for all criteria doesn't reflect real-world importance
• No penalty for very long or very short descriptions
• Missing checks for: example code quality, API coverage completeness

Ideas

• Weighted scoring: description (25%), examples (25%), API docs (20%), triggers (15%), references (15%)
• Penalize generic/templated descriptions
• Check that code examples are syntactically valid
• Compare against high-quality skills on ClawHub for benchmarking
• Add --explain-score\ flag to show breakdown

Good starter task — the scoring logic is in \src/quality.ts.