What is Auto-Skill?
Auto-Skill is an SDK library that automatically learns from your coding workflows and transforms them into intelligent, context-aware skills. It observes your tool usage patterns, detects repetition, and proactively suggests both local patterns and community skills from 27,000+ external sources before you even ask. Instead of manually creating skills or searching for solutions, Auto-Skill works in the background to:- Observe your coding sessions and capture workflow events
- Detect repeated patterns and analyze your problem-solving approaches
- Suggest relevant community skills or generate custom skills from your patterns
New in v5.0: Proactive Skill Discovery combines local pattern detection with external skill discovery, creating a hybrid learning system that suggests battle-tested community skills matched to your detected workflows.
Key Features
Proactive Discovery
Context-aware recommendations that automatically suggest community skills based on detected frameworks, languages, and intent. Search 27,000+ skills from Skills.sh in real-time.
Pattern Detection
Automatically detects repeated tool sequences across sessions. Identifies 3+ occurrences with high confidence scoring and tracks success rates.
Session Analysis
Identifies intent (debug, implement, refactor, test) and workflow types (TDD, deployment, etc.). Recognizes 18 design patterns including architectural, coding, and workflow patterns.
Multi-Agent Support
Cross-agent skill sharing for Claude Code, Codex, Cursor, Aider, Windsurf, and more. Skills work across multiple AI coding agents via symlinks.
Smart Graduation
Suggests upgrading local patterns to battle-tested community skills. Confidence evolves from 50% (external) → 75% (proven) → 85% (graduated).
Provider System
Pluggable skill discovery with local, Skills.sh, and RFC 8615 well-known endpoint providers. Extensible architecture for custom sources.
How It Works
Auto-Skill uses a three-stage pipeline architecture:1. Observer (Event Capture)
Hooks into your coding agent’s tool execution flow to capture workflow events including tool calls, outcomes, and context. Events are stored in a local SQLite database.2. Detector (Pattern Recognition)
Analyzes captured events for reusable patterns. Detection triggers include:- Repetition: Same tool sequence 3+ times
- Success patterns: High success rate on similar tasks
- User corrections: When you redirect the agent’s approach
- Explicit teaching: When you provide domain knowledge
- Error recovery: Successful self-correction patterns
3. Skill Forge (Generation)
Generates valid SKILL.md files from detected patterns or loads community skills from external sources. All skills follow the agentskills.io specification.Auto-Skill uses a human-in-the-loop model. Skills require your confirmation before activation, ensuring you maintain control over what gets added to your workflow.
Use Cases
- Framework Setup
- Testing Workflows
- Deployment
- Debugging
Automatically detect when you’re setting up React, Next.js, or other frameworks repeatedly. Auto-Skill suggests community skills for scaffolding, configuration, and best practices.
Quick Links
Installation
Install the SDK and CLI
Quick Start
Get up and running in minutes
Core Concepts
Learn the architecture
Why Auto-Skill?
Traditional approaches require you to:- Manually search for skills
- Remember to apply patterns you’ve used before
- Recreate solutions you’ve already built
- Skills find you based on your context
- Patterns are automatically detected and codified
- Community knowledge is proactively suggested when relevant