@skillbolt/tree

Capability tree builder for Skillbolt. Organizes skills into a hierarchical tree structure using LLM-powered classification.

Features

• LLM-powered categorization — Automatically assigns skills to categories using LLM
• Configurable branching factor — Control tree width and depth
• Parallel BFS building — Queue-based breadth-first construction with p-limit concurrency
• Multiple output formats — YAML, JSON, ASCII, HTML serialization
• Fixed root categories — Predefined top-level categories for consistency

Quick Start

import { TreeBuilder, createTreeConfig, loadTree } from '@skillbolt/tree';
import { createLLMAdapter } from '@skillbolt/core';

const llm = createLLMAdapter({ model: 'gpt-4o-mini' });
const config = createTreeConfig({ branchingFactor: 8, maxDepth: 6 });
const builder = new TreeBuilder({ llm, config });

const result = await builder.build({
  skillDir: '.claude/skills',
  onProgress: (p) => console.log(p.message),
});

// Load a previously built tree
const tree = await loadTree('tree.yaml');
console.log(tree.countAllSkills());

CLI

skill tree build --dir .claude/skills --verbose
skill tree show --format ascii
skill tree stats

API

| Export | Description | | ------------------ | -------------------------------------------------- | | TreeBuilder | Main builder class (queue-based BFS with LLM) | | TreeNode | Tree node with children, skills, traversal methods | | createTreeConfig | Create config with derived properties | | loadTree | Load tree from YAML file | | renderTreeASCII | ASCII visualization | | saveTreeToJSON | JSON serialization | | saveTreeToHTML | HTML visualization | | getTreeStats | Node count, skill count, depth |

Configuration

| Option | Default | Description | | ----------------- | -------- | ------------------------ | | branchingFactor | 8 | Max children per node | | maxDepth | 6 | Maximum tree depth | | maxWorkers | 4 | Concurrent LLM calls | | rootCategories | Built-in | Top-level category names |

License

MIT