# skill-tree-architect Whole-tree architect for a skill library's hub-and-spoke taxonomy — audits description-cap headroom, hub balance, and cross-hub placement, then rebalances for a new family. Every other skill on this page adds to a skill library one piece at a time — [concept-family-explorer](/skills/concept-family-explorer/) and [rabbithole](/skills/rabbithole/) fill in research, [full-suite](/skills/full-suite/) compiles it into llms-family files. None of them look at the *shape* of the resulting tree. `skill-tree-architect` does: it audits tree-wide description-cap headroom, hub balance, and cross-hub placement, then drives the rebalancing toolchain to reshape the tree around a newly-grown family. It runs read-only analysis first — placement audit, candidate detection, structural lint — and surfaces a ranked rebalance plan: which hubs are over their description cap, which spokes are filed under the wrong hub, which standalone skills need a hub of their own. Zero-risk, idempotent repairs get applied directly; folding, splitting, and registry sync get surfaced for human review rather than applied silently. **Use it for:** "rebalance the skill tree", "is my taxonomy optimal", "the skills folder is a mess, reorganize it", "find misplaced spokes". **Not for:** one skill's own content, triggers, or peer-seeds (that's a single-skill audit, not a tree-shape problem) · deciding what to build next (concept-family-explorer) · a one-off structural lint of a single file.