SkillKit

Self-customizing workflow orchestration for AI-assisted development.

Granular, doc-based workflows that adapt to your project + Anthropic skills integration + Community marketplace.

A cross-platform system that combines intelligent workflows, terminal-aware skill loading, and self-customization into one powerful tool for systematic development.

🎯 What's New

SkillKit is a self-customizing workflow orchestration system focused on doc-based workflows:

✨ Key Features:

• ✅ 13 Production Workflows - BEGIN_SESSION, IMPLEMENT_FEATURE, FIX_BUGS, SKILLKIT_TASK, and more
• ✅ Intelligent Task Planning - tsk plan and tsk task for automatic skill selection and execution
• ✅ Auto-Customization - Workflows automatically adapt to your project during tsk init
• ✅ Self-Customizing - META workflows adapt to your project automatically
• ✅ Terminal-Aware - Cross-platform skill loading (Windows/Mac/Linux)
• ✅ Community Marketplace - Install workflows/skills from GitHub with one command
• ✅ System Auditing - Comprehensive health checks with auto-fix
• ✅ Usage Telemetry - Track skill usage and performance with tsk stats
• ✅ Anthropic Skills - Integrated PDF, Excel, database expertise on-demand

🚀 IDE Support Status:

• ✅ Cursor: Full support with slash commands (/BEGIN_SESSION, /IMPLEMENT_FEATURE)
• 🔜 VS Code: CLI + tasks.json (coming soon)
• 🔜 Other IDEs: Universal AGENTS.md support (roadmap)

What Makes SkillKit Unique?

The Self-Customizing Workflow System:

🔄 Hierarchical Workflows

• Main Workflows - Multi-step procedures (/IMPLEMENT_FEATURE, /FIX_BUGS)
• Subtasks - Reusable components (22 granular subtasks)
• Skills - Domain expertise (PDF, Excel, databases via Anthropic)
• Hierarchical - Like functions in programming, workflows call subtasks call skills

🎯 Auto-Customization (The Moat!)

• Automatic During Init - Workflows adapt to YOUR project during tsk init
• Intelligent Detection - Detects languages, package managers, test frameworks, linters, formatters
• Smart Adaptation - All workflow commands updated to match your environment
• Manual Override - Use --no-auto-customize to skip, or /META_CUSTOMIZE for fine-tuning
• Evolves - REVIEW_SKILLKIT continuously improves based on your usage

🌐 Terminal-Aware Cross-Platform

• Smart Execution - Detects shell (PowerShell/CMD/Bash/Zsh)
• Auto-Translates - Commands work everywhere (Windows/Mac/Linux)
• Skill Loading - tsk skill:load pdf works on any platform
• No Configuration - Just works™

🏪 Community Marketplace

• One Command Install - tsk skills:add user/repo/skill-name
• GitHub-Based - Free, distributed, no central server
• Auto-Validate - Checks format before installing
• Easy Share - Publish to GitHub, others install instantly

Installation

# Using npm
npm install @trinity-os/skillkit

# Using yarn
yarn add @trinity-os/skillkit

# Using pnpm
pnpm add @trinity-os/skillkit

Quick Start

🚀 Automatic Setup (Recommended)

# Install in your project
cd your-project
npm install @trinity-os/skillkit

# That's it! SkillKit auto-initializes during install:
# ✓ Detects your project (package.json, .git, etc.)
# ✓ Copies 13 workflows to .cursor/commands/ (including SKILLKIT_TASK)
# ✓ Auto-customizes workflows to your project (detects languages, package managers, tools)
# ✓ Installs Anthropic skills (pdf, xlsx, docx, etc.)
# ✓ Generates AGENTS.md catalog
# ✓ Auto-deduplicates any conflicts
# ✓ Ready to use immediately!

Note: To skip auto-initialization, set SKILLKIT_NO_AUTO_INIT=true before install.

🛠️ Manual Initialization (Optional)

# If you need to manually initialize or re-initialize
tsk init --cursor

# Or for non-Cursor projects
tsk init --workflows

📋 Use Workflows in Cursor

Type "/" in Cursor to see available workflows:

/BEGIN_SESSION          → Start session (shows menu!)
/IMPLEMENT_FEATURE      → Build new feature
/FIX_BUGS              → Fix bugs systematically
/DEPLOY_PREP           → Pre-deployment checks
/SKILLKIT_TASK         → Unified task execution (uses tsk plan/task)
/CONTINUE              → Resume from last session
/AUDIT_SKILLKIT        → System health check
/SECURITY_AUDIT        → Security scan
/META_CUSTOMIZE        → Adapt to your project
/HELP                  → Complete documentation

🎯 Auto-Customization (Automatic!)

During tsk init:
✓ Automatically detects your languages (TypeScript, Python, etc.)
✓ Detects package managers (npm/pnpm/yarn/poetry/pip)
✓ Detects tools (ESLint, Prettier, Jest, pytest, etc.)
✓ Adapts ALL workflow templates to YOUR project
✓ Commands like "npm test" become "pnpm test" everywhere

No manual step needed - it just works!

🧠 Intelligent Task Execution

# Plan which skill to use for a task
tsk plan "fix all ESLint errors"
# → Analyzes available skills
# → Selects best match with confidence score
# → Shows reasoning

# Execute task through SkillKit
tsk task "fix all ESLint errors"
# → Plans automatically
# → Executes selected skill
# → Logs usage for statistics

# View usage statistics
tsk stats
# → Shows skill usage counts
# → Average execution times
# → Success/failure rates

🎯 Manual Customization (Optional)

In Cursor:
/META_CUSTOMIZE

For advanced customization:
✓ Fine-tune workflow steps
✓ Add project-specific rules
✓ Customize command mappings

📦 Skills Management

# Install Anthropic skills (done automatically in init)
tsk install anthropics/skills
# → Interactive checkbox to select skills
# → Installs to .claude/skills/

# Load a skill when needed (MANDATORY for agents)
tsk skill:load pdf
# → Outputs full SKILL.md content (200-600 lines) to terminal
# → AI agent MUST read terminal output to get skill knowledge
# → Works cross-platform (Windows/Mac/Linux)
# → Skills are loaded into context via terminal output

# List installed skills
tsk list

# Install community skills
tsk skills:add alice/db-skills/postgres
# → Auto-downloads, validates, installs

⚠️ Important for AI Agents:

• Skills are NOT automatically loaded - you MUST run tsk skill:load <name> in terminal
• The terminal output contains the full skill instructions
• You MUST read the terminal output to get skill knowledge
• Workflows explicitly require skill loading via terminal commands

🛠️ System Commands

# Intelligent task planning and execution
tsk plan "<task description>"     # Plan which skill to use
tsk task "<task description>"      # Execute task (plan + run)
tsk stats                          # View skill usage statistics

# Run diagnostics
tsk diagnose
# → Auto-detects project type
# → Runs lint, typecheck, tests
# → Shows issues

# System audit
tsk audit
# → Checks for duplicates
# → Validates workflows
# → Tests commands
# → Generates health score

# Auto-fix safe issues
tsk audit:fix --auto-safe

# Code quality checks
tsk run-checks
# → Runs lint, typecheck, build
# → Checks circular dependencies (madge)
# → Generates comprehensive report

# Sync AGENTS.md
tsk sync

🏪 Community Marketplace

# Install community workflows
tsk workflows:add john/devops-workflows/DEPLOY_K8S.md
# → Downloads from GitHub
# → Installs to .cursor/commands/
# → Available as /DEPLOY_K8S

# Install community skills
tsk skills:add alice/enterprise-skills/database
# → Downloads from GitHub
# → Installs to .claude/skills/
# → Updates AGENTS.md

Documentation

Getting Started

• Type /HELP in Cursor - Complete interactive documentation
• Workflow System Explained - How CLI, workflows, and subtasks work
• Current Status - What's implemented
• Marketplace & Contribution - Community guide

Architecture & Vision

• Final Correct Architecture - The complete system
• Cursor-First Roadmap - Development roadmap
• Skill Update Strategy - How updates work
• Research Enhancements - Academic research integration

📚 Research & Attribution

SkillKit incorporates insights from academic research papers on skill learning, task planning, and agent systems:

Research Papers Integrated

1. SkillFlow: Efficient Skill and Code Transfer Through Communication
   arXiv: 2504.06188

   • Modular skill acquisition framework
   • Communication-based skill transfer
   • Applied: Skill composition and dynamic discovery

2. COALESCE: Economic and Security Dynamics of Skill-Based Task Outsourcing
   arXiv: 2506.01900

   • Dynamic skill discovery
   • Automated task decomposition
   • Cost-aware skill selection
   • Applied: Task decomposition and cost-aware planning

3. PolySkill: Learning Generalizable Skills Through Polymorphic Abstraction
   arXiv: 2510.15863
   Authors: Simon Yu, Gang Li, Weiyan Shi, Peng Qi

   • Polymorphic abstraction for skill generalization
   • Skill composition and reuse
   • Applied: Skill abstraction layer and composition support

4. SkillWeaver: Web Agents can Self-Improve by Discovering and Honing Skills
   arXiv: 2504.07079
   Authors: Boyuan Zheng, Michael Y. Fatemi, Xiaolong Jin, et al.

   • Self-improvement through skill discovery
   • Skill refinement and optimization
   • Applied: Self-improvement mechanisms and usage-based optimization

License Note: All papers are from arXiv and are typically available under licenses that permit commercial use (CC-BY or similar). We recommend verifying specific license terms for each paper before commercial deployment.

Full Details: See Research Enhancements for complete implementation details.

For Contributors

• Contributing Guidelines - How to contribute
• Marketplace Guide - Publish your workflows/skills
• Topic tags: skillkit-workflow, skillkit-skill on GitHub

Community & Marketplace

Find Community Content

Search GitHub for:

• Topic: skillkit-workflow - Community workflows
• Topic: skillkit-skill - Community skills
• Or browse: Awesome SkillKit (coming soon)

Share Your Work

# 1. Create repo with your workflows/skills
# 2. Add topic: skillkit-workflow or skillkit-skill
# 3. Share in GitHub Discussions
# 4. Others install with: tsk workflows:add your-username/repo

Get Help

• GitHub Discussions - Ask questions
• Issues - Report bugs
• Type /HELP in Cursor - Complete documentation

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• Thanks to all our contributors who have helped make this project better.
• Special thanks to our early adopters and beta testers.

Sponsors

Support this project by becoming a sponsor. Your logo will show up here with a link to your website. [Become a sponsor]

Contributors ✨

Thanks goes to these wonderful people:

Jai Thakur 💻 🚇 🐛

This project follows the all-contributors specification. Contributions of any kind welcome!