SkillDB API & Integration Guide

# Install and initialize
npm install -g skilldb
skilldb init

# Search for the skill you need
skilldb search react hooks

# Get exactly the skill you need
skilldb get react-patterns-skills/custom-hooks

# Your agent reads it from .skilldb/skills/ — zero latency

# CLAUDE.md

## Skills
This project uses SkillDB skills cached in .skilldb/skills/.
Read skill files from that directory before starting tasks.

To add more skills: skilldb search <query> && skilldb get <pack>/<skill>
Full index: https://skilldb.dev/api/v1/skills

skilldb/
  skills-data.json          # Full index (categories, packs, skill metadata)
  packs/
    author-styles/
      hemingway.md           # Individual skill file
      orwell.md
    software-skills/
      react-architect.md
      api-design.md
    ...
  web/                       # This Next.js app
    src/app/
      api/skills/route.ts    # REST API
      api/stats/route.ts

---
title: "React Architect"
description: "Senior React architect specializing in scalable component design"
category: "Technology & Engineering"
pack: "software-skills"
type: "functional"           # functional | person-style | critic-style
version: "1.0"
tags: ["react", "architecture", "frontend"]
---

---
title: "Ernest Hemingway"
description: "Write with Hemingway's iconic minimalist prose style"
category: "Writing & Literature"
pack: "author-styles"
type: "person-style"
version: "1.0"
tags: ["writing", "fiction", "minimalism"]
---

# Ernest Hemingway

## Identity
You write in the style of Ernest Hemingway. Your prose is
spare, direct, and emotionally resonant through what is
left unsaid.

## Principles
- Use short, declarative sentences
- Prefer concrete nouns and active verbs
- Omit needless words — apply the Iceberg Theory
- Let dialogue carry the emotional weight
- Avoid adverbs and adjectives where the verb suffices

## Anti-patterns
- Do NOT use flowery or ornate language
- Do NOT explain emotions — show them through action
- Do NOT use passive voice unless deliberately

# CLAUDE.md

## Skills
Use SkillDB skills from https://skilldb.dev/api/v1/skills
When a task matches a skill category, fetch and apply the relevant skill.

## Active Skills
- author-styles/hemingway — for blog posts
- software-skills/react-architect — for component work

# .cursorrules

You have access to SkillDB skills.
Skills index: https://skilldb.dev/api/v1/skills

When writing React code, apply the react-architect skill.
When writing API routes, apply the api-design skill.
Fetch the skill markdown and follow its instructions.

# Initialize and get a skill for offline use
skilldb init
skilldb get author-styles/hemingway

# Skills are cached in .skilldb/skills/ for zero-latency reads
# Reference in codex instructions
echo "Apply the skill in .skilldb/skills/author-styles/hemingway.md" >> instructions.md

// Fetch the skills index
const res = await fetch("https://skilldb.dev/api/v1/skills");
const { skills } = await res.json();

// Find skills matching a task
const writingSkills = skills.filter(
  s => s.category === "Writing & Literature"
);

// Load a specific skill's content
const skill = await fetch(
  "https://skilldb.dev/api/v1/skills/author-styles/hemingway"
);
const { content } = await skill.json();

// Inject into your agent's system prompt
const systemPrompt = `${basePrompt}\n\n## Active Skill\n${content}`;

# Use directly with npx (no install needed)
npx skilldb search react hooks
npx skilldb get react-patterns-skills/custom-hooks

# Or install globally for the full workflow
npm install -g skilldb

# Initialize SkillDB in your project
skilldb init

# Search → Get → Done
skilldb search code review
skilldb get software-skills/code-review

# Search skills by keyword (multi-word search supported)
skilldb search react hooks
skilldb search code review
skilldb search api design --category "Technology & Engineering"

# Get exactly the skill you need (recommended)
skilldb get react-patterns-skills/custom-hooks
skilldb get software-skills/code-review
skilldb get autonomous-agent-skills/self-correction

# Download an entire pack (when you want everything)
skilldb add software-skills
skilldb add autonomous-agent-skills

# List available categories and skills
skilldb list
skilldb list --category "Autonomous Agents"
skilldb list --pack software-skills

# Show skill details and preview
skilldb info software-skills/code-review

# Save your API key (for full content access)
skilldb login

# Remove inactive cached skills to free disk space
skilldb purge
skilldb purge --all          # remove everything including active
skilldb purge --inactive     # only remove unused skills
skilldb purge --dry-run      # preview what would be removed

import { createClient } from 'skilldb';

const db = createClient(); // auto-loads key from env/config

// Search for skills
const results = await db.search('code review');
console.log(results.skills);

// Get a single skill with full content
const skill = await db.get('software-skills/code-review.md');
console.log(skill.title, skill.content);

// List with filters
const listing = await db.list({
  category: 'Autonomous Agents',
  limit: 10,
});

import { createClient, initCache, cacheSkill } from 'skilldb';

const db = createClient();
initCache(); // creates .skilldb/ directory

// Download and cache a skill locally
const skill = await db.get('software-skills/code-review.md');
const path = cacheSkill(skill);
console.log('Cached to:', path);

# 1. Environment variable (CI-friendly)
export SKILLDB_API_KEY=sk_live_xxx

# 2. Project-level config
echo '{"apiKey":"sk_live_xxx"}' > .skilldbrc

# 3. User-wide config (set via skilldb login)
# Saved to ~/.skilldbrc

# Step 1: Install globally (one time)
npm install -g skilldb

# Step 2: Add to Claude Code
claude mcp add skilldb -- skilldb-mcp

# With API key (for full skill content)
claude mcp add skilldb -- skilldb-mcp --api-key sk_live_xxx

# Cursor — add to .cursor/mcp.json
{
  "mcpServers": {
    "skilldb": {
      "command": "skilldb-mcp"
    }
  }
}

# Windsurf — add to mcp_config.json
{
  "mcpServers": {
    "skilldb": {
      "command": "npx",
      "args": ["skilldb-mcp", "--api-key", "sk_live_xxx"]
    }
  }
}

# Any MCP client
skilldb-mcp

# In Claude Code or Cursor, just ask:

"Search SkillDB for React performance patterns"
→ Agent calls skilldb_search("react performance")
→ Returns matching skills with descriptions

"Load the code review skill"
→ Agent calls skilldb_get("software-skills/code-review.md")
→ Full skill content loaded into context

"What skills should I use for this project?"
→ Agent calls skilldb_recommend with your tech stack
→ Returns tailored pack recommendations

"Show me all security skills"
→ Agent calls skilldb_list(category: "Security")
→ Lists all available security skill packs

"Clean up unused skills from my project"
→ Agent calls skilldb_purge(mode: "inactive")
→ Removes cached skills not in active profile