The skills dataset moved into a tree (skills/core, skills/specs, ...).
Resolve it data-driven instead of by name:
- The required-family check no longer pins skill focus families to
dataset "skills"; the family spec's child_source names the dataset.
- Affinity generation and display names use the skill_focus spec's
child_source dataset.
- Child slugs derive from the row key's own namespace (text after the
first colon) instead of assuming namespace == dataset path.
- The wiki loader accepts both layouts ("skills", then "skills/core").
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>