Skip to content

X-School-Academy/skill-pilot

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

602 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Skill Pilot AI Agent

Your personal AI workspace — learn, automate, and build from a single idea.

Software runs. Codeware grows.

Whether you are just starting out with AI, a professional who wants to automate your daily work, or a team building a full commercial product — Skill Pilot gives you one unified environment powered by AI agents.

Join the community:

💬 Discord | 𝕏 | Reddit | WebChat

Installation

Supported platforms: macOS, Linux, Windows (via WSL only)

Supported AI agents: Claude Code, GitHub Copilot CLI, Codex, Gemini CLI, OpenCode — or enable Ollama, OpenAI-compatible, and Claude-compatible providers in config/ai_providers.json5

Auto Install

curl -fsSL https://skill-pilot.ai/install.sh | bash

Manual Install

Follow the step-by-step guide in SETUP.md.

Media MCP

For media generation and speech workflows support, you need to setup Skill Pilot Media docker separately:

Supported model families include Wan2.2, Z-Image, IndexTTS, SongBloom, and others.

Then see Usage to get started.


Intro Video

Intro


Who Is This For?

🌱 Beginners — Learn AI from Zero to Hero

You have heard about AI and vibe coding but don't know where to start. Skill Pilot gives you:

  • Interactive AI courses generated just for you — no prior coding experience needed
  • Vibe coding — describe what you want in plain English and watch AI build it
  • Hands-on projects — spin up real working apps without writing a single line of code yourself
  • A structured path from complete beginner to confident AI user

Start here: open the workspace/learning/ folder and ask the AI to create a course on any topic.


⚡ Professionals — Automate Your Daily Work with AI

You already know your domain. Now you want AI to handle the repetitive, time-consuming work — not just in code, but across everything you do. Skill Pilot gives you:

  • Task automation — delegate writing, research, analysis, planning, and scheduling to AI agents
  • Computer control — let AI operate your desktop, fill forms, extract data from websites, and interact with any app
  • Remote access — control servers and cloud infrastructure through natural language
  • Discord and communication bots — AI-powered assistants that work while you sleep
  • A growing library of reusable skills you can trigger anytime

Start here: open the workspace/tasks/ folder and describe any task you want automated.


🏢 Enterprises — Build a Commercial Product from a Single Idea

You have an idea and want to go from concept to a fully deployed, commercial-ready product. Skill Pilot's Dev Swarm system handles the entire software lifecycle:

  1. Idea → Requirements — AI clarifies and structures your business problem
  2. Market Research — competitive analysis and opportunity identification
  3. Architecture & Tech Specs — system design, API specs, database schema
  4. Sprint Planning & Development — feature-by-feature implementation with tests
  5. DevOps & Deployment — CI/CD, containerization, cloud infrastructure
  6. Monitoring & Iteration — production-ready and continuously improving

Start here: write your idea in ideas.md and ask Dev Swarm to begin stage 1.


🎯 Everyone — Master AI and Prepare for the AI Era

The most valuable skill in the next decade is knowing how to work with AI effectively. Skill Pilot is designed to grow with you:

  • Agent skills you build become reusable assets — your AI gets smarter over time
  • Community knowledge sharing — learn from and contribute to a growing library of capabilities
  • Multi-agent workflows — chain AI agents together to solve complex, multi-step problems
  • Your personal AI habitat — a workspace that evolves as you do

What Is Skill Pilot?

Skill Pilot is an AI-native workspace you run locally. It connects your favorite AI coding agents (Claude Code, GitHub Copilot CLI, Codex, Gemini CLI, and more) to a structured environment of reusable skills, tools, and workflows.

Instead of using AI as a chat assistant, Skill Pilot turns AI into an active worker that:

  • Reads and modifies your codebase
  • Executes tasks on your computer
  • Learns new capabilities through reusable "skills"
  • Operates across your entire digital workflow — not just code

In short: You describe what you want. The AI does it. The system gets better every time.


What Is Codeware?

Traditional software is static — it ships with fixed features that only change when developers update it.

Codeware is different. The source code itself is the product, and AI continuously extends it based on what you need.

Traditional:  Human → UI → Software → Fixed behavior
Codeware:     Human → AI → Codebase → Evolving behavior

In Skill Pilot:

Traditional Software Skill Pilot Codeware
You use features You grow capabilities
Updates come from developers Updates come from AI + you
Fixed at release Evolves through use
One-size-fits-all Adapts to your needs

The system is not "used". It is grown.


Usage

After installation, run:

# cd <your installation path>
./skillpilot.sh help
Usage: ./skillpilot.sh [help|build|start|stop|test|doctor] [--dev]
       ./skillpilot.sh <enable|disable> <human-detection|live-tts>
       ./skillpilot.sh doctor [question]

Commands:
  help    Show this help message.
  build   Build static webui export (core/webui/www).
  start   Start services. Default command.
  stop    Stop running tmux sessions. Use `--dev` to stop only development sessions.
  test    Run engine pytest suite, or only the named files under core/engine/tests.
  doctor  Open the configured doctor AI agent for troubleshooting and guided help.
  enable human-detection    Install optional human detection dependencies.
  disable human-detection   Uninstall optional human detection dependencies.
  enable live-tts           Install optional live-tts dependencies.
  disable live-tts          Uninstall optional live-tts dependencies.

Options:
  --dev   Use development mode for `start`, or stop only development sessions for `stop`.
  --next-server
          Production mode only. Serve the WebUI via `next start` (the Next.js production server)
          instead of the static HTML export served by the engine. For testing the production
          build of the WebUI. Cannot be combined with `--dev`.

Defaults:
  - Command defaults to: start
  - Mode defaults to production (without --dev)

Prod and Dev Modes

Skill Pilot now supports running production and development side by side:

  • Production engine and bundled release WebUI: http://127.0.0.1:3001
  • Development engine: http://127.0.0.1:3002
  • Development WebUI: http://127.0.0.1:3003

Typical workflow:

./skillpilot.sh
./skillpilot.sh --dev

Stop commands:

./skillpilot.sh stop
./skillpilot.sh stop --dev
  • ./skillpilot.sh stop stops production only
  • ./skillpilot.sh stop --dev stops development only

Testing the production WebUI build (--next-server)

For testing the production build of the Next.js WebUI without the static export served by the engine, start prod mode with --next-server:

./skillpilot.sh start --next-server
./skillpilot.sh stop  --next-server

This runs the engine on its prod port and additionally starts next start (Next.js production server) on the WebUI port (default 127.0.0.1:3003). Use this only for local testing of the production WebUI bundle — normal usage should stick with ./skillpilot.sh (static export served by the engine).

Using AI Agent CLIs Directly

If you use agent CLIs directly, without the Skill Pilot WebUI:

  • For normal usage, start the engine in production mode with ./skillpilot.sh
  • For development workflows, export SKILL_PILOT_RUNTIME_MODE=development before starting the agent CLI directly

Examples:

export SKILL_PILOT_RUNTIME_MODE=development
claude
export SKILL_PILOT_RUNTIME_MODE=development
codex
export SKILL_PILOT_RUNTIME_MODE=development
gemini

This ensures the agent CLI can see the same runtime mode as the development engine.

Troubleshooting with ./skillpilot.sh doctor

If you get stuck during setup or normal usage, you can open the built-in doctor agent:

./skillpilot.sh doctor

You can also pass the problem directly on the command line:

./skillpilot.sh doctor "install.sh failed while installing dependencies"
./skillpilot.sh doctor "skillpilot.sh start does not open the WebUI"

What it is useful for:

  • Troubleshooting install.sh or skillpilot.sh
  • Explaining technical terms or error messages in plain language
  • Finding likely environment or dependency problems on your machine
  • Investigating bugs in the Skill Pilot codebase

Behavior:

  • If no question is provided, Skill Pilot will prompt you and show example questions
  • It will try to use the configured doctor provider from config/ai_providers.json5
  • If the doctor provider is opencode and it is not installed yet, Skill Pilot will try to install it automatically
  • Once the doctor session starts, you can chat with it like a normal AI assistant
  • Exit the doctor session with /exit or double Ctrl-C

The doctor agent is designed for beginners and can inspect project docs, skills, and source code to help diagnose issues. If a real code fix is needed, it should ask for your approval before making code changes.

For beginner-friendly examples of how to write a good doctor request, see Skill Pilot Doctor Message Guide.


Your Workspace

After installation, your workspace is organized into focused zones — each powered by Skill Pilot.

skill-pilot/
├── src/                    ← Your commercial-ready product
│   └── ...                    Built end-to-end by Dev Swarm agents
│
└── workspace/
    ├── learning/           ← Your AI study studio
    │   └── ...                Courses, tutorials, and interactive lessons
    │
    ├── projects/           ← Your vibe coding lab
    │   └── ...                Spin up any coding project — fast
    │
    ├── research/           ← Your research workspace
    │   └── ...                Deep dives, analysis, and knowledge gathering
    │
    └── tasks/              ← Your AI task runner
        └── ...                Automate any non-coding task with AI
Zone What it is Best for
src/ The product you are building — from idea to deployment, managed by Dev Swarm Enterprises, product builders
workspace/learning/ AI-powered courses, tutorials, and skill-building Beginners, learners
workspace/projects/ Vibe coding space — describe what you want, watch it get built Beginners, developers
workspace/research/ Research workspace for analysis, competitive research, and knowledge synthesis Professionals, researchers
workspace/tasks/ Task runner for everything else — writing, automation, planning, and more Professionals, everyone

Features

Progress indicator:

Symbol Status
⭐⭐⭐⭐⭐ Ready to use
⭐⭐⭐⭐ Ready for test
⭐⭐⭐ Code is ready
⭐⭐ Wait to open source
Planned

Core Platform

Feature Status
Skill Pilot Core Engine — Python FastAPI backend with LLM routing, session management, and MCP servers ⭐⭐⭐⭐⭐
Skill Pilot WebUI — Next.js interface for chat, terminal, workflow editor, course viewer, and settings ⭐⭐⭐⭐⭐
Multiple AI Code Agent Support — Claude Code, GitHub Copilot CLI, Codex, Gemini CLI, OpenCode CLI ⭐⭐⭐⭐⭐
LLM API Support — Ollama, OpenAI-compatible, and Claude-compatible endpoints via config/ai_providers.json5 ⭐⭐⭐⭐⭐
Background Process Management — tmux-based session lifecycle for long-running agents ⭐⭐⭐⭐⭐
WebUI Terminal — Full browser-based terminal via WebSocket ⭐⭐⭐⭐⭐

Agent Skills & Automation

Feature Status
Agent Skill Creator — create, update, and find reusable agent skills via agent-skill ⭐⭐⭐⭐⭐
Import Agent Skill by Git URL — install third-party skills directly from any git repository ⭐⭐⭐⭐⭐
Background Prompt via background-prompt — execute a task in the background with a prompt through the core engine ⭐⭐⭐⭐⭐
New Agent Session via new-agent-session — spawn a new agent process in an existing terminal session ⭐⭐⭐⭐⭐
Agent Skill Schedule Support — cron-based skill scheduling via APScheduler ⭐⭐⭐⭐⭐
MCP to Skills — expose MCP server tools as callable agent skills ⭐⭐⭐⭐⭐
Multiple Agent Workflow Diagram Creator & Executor — visual multi-agent workflow editor and runner ⭐⭐⭐⭐

Development Tools & Extensions

Feature Status
VS Code Extension — IDE integration for Skill Pilot agent control ⭐⭐⭐⭐⭐
Remote SSH Access via terminal MCP — full remote terminal, SCP file transfer, and SSH tunneling ⭐⭐⭐⭐⭐
Dev Swarm — full software lifecycle from ideas to deployment for building commercial-ready products ⭐⭐⭐⭐
Chrome Extension — browser integration for Skill Pilot ⭐⭐⭐⭐

Security

Feature Status
KeySafe Guard via keys-safe-guard skill — protect and manage LLM API keys in config/.env ⭐⭐⭐⭐⭐

AI Media Generation (via media MCP)

Feature Status
ComfyUI Image Generation — text-to-image and image-to-image via ComfyUI workflows ⭐⭐⭐⭐⭐
ComfyUI Video Generation — text-to-video, image-to-video, and video-to-video workflows ⭐⭐⭐⭐⭐
ComfyUI LipSync — talking head video from image or video with audio ⭐⭐⭐⭐⭐
Musetalk Live Avatar — real-time AI avatar powered by MuseTalk ⭐⭐⭐⭐⭐
MuesTalk LipSync — lip-sync video generation via MuseTalk CLI ⭐⭐⭐⭐⭐
IndexTTS2 — high-quality speech synthesis via IndexTTS2 ⭐⭐⭐⭐⭐
SongBloom AI Music Creator — AI music generation with vocals and accompaniment ⭐⭐⭐⭐⭐
AI Agent Live TTS via live-tts MCP — real-time text-to-speech for agent responses ⭐⭐⭐⭐⭐
GenAI Image Support — generative image tools available to agents ⭐⭐⭐⭐
GenAI TTS Support — generative text-to-speech tools available to agents ⭐⭐⭐⭐

Education & Courses

Feature Status
Personal Course Creator — AI-generated interactive courses via LangGraph workflow ⭐⭐⭐⭐⭐
Tutorial Creator with Realtime Code Execution and AI Assistant — live in-browser code runner with AI feedback ⭐⭐⭐⭐⭐
Text / Image / Audio / Video Creator — multi-modal content pipeline via LangGraph workflow ⭐⭐⭐⭐⭐

Desktop Automation & Communication

Feature Status
Screen Recording Skill — record the macOS screen from the CLI via tmux and ffmpeg ⭐⭐⭐⭐⭐
Screen Drawing Skill — draw temporary bounding-box overlays on macOS for UI highlighting ⭐⭐⭐⭐⭐
Discord Bot Support — multi-session Discord bot with LLM-powered AI assistant ⭐⭐⭐⭐⭐
Use Computer Skill — GUI automation via screenshots, mouse control, and keyboard input ⭐⭐⭐⭐

Smart Home / IoT

Feature Status
Local Security Cameras ⭐⭐
Z-Wave Smart Home ⭐⭐
E2EE Security Camera Hub ⭐⭐
Security Camera Firmware / Cloud Code / Mobile App ⭐⭐

How Skill Pilot Works

  1. You run Skill Pilot locally on your machine
  2. The AI reads your repository and understands your project
  3. You describe what you want — in plain English
  4. The AI edits code, runs tasks, or creates new skills to solve it
  5. New capabilities are saved as reusable skills for next time
  6. Useful skills can be shared back to the community via contrib
  7. Stable, proven skills are released into codeware for everyone

The system improves every time it is used.


Branch Model

Skill Pilot separates three layers of evolution so your personal work never conflicts with shared releases.

Branch Purpose Who writes
codeware Stable release layer Maintainers
contrib Shared improvements between users Community + AI review
user Your personal evolving workspace You + your AI

Flow of knowledge

codeware  → user        (receive updates)
user      → contrib     (share what works)
contrib   → codeware    (stabilized release)

Rules:

  • codeware is protected — release only
  • contrib accepts pull requests from the community
  • user is freely modified by you and your AI
  • AI never edits codeware directly

Contributing

You do not contribute by writing features for users. You contribute by teaching the system a reusable skill.

Steps:

  1. Work in your user branch
  2. If multiple users share one computer, each user should fork to a different repo and clone locally to a different path
  3. When something is useful → create a PR to contrib
  4. Maintainers and AI review integration safety
  5. Merged into shared knowledge
  6. Released into codeware

Philosophy

Software should not contain all knowledge. It should learn knowledge.

Traditional software distributes features. Codeware distributes capability evolution.

You are not installing an application.

You are installing a habitat for an AI worker.

The AI does not operate inside the program. The program operates inside the AI.


Release Concept

Releases are not feature milestones. They are intelligence stabilization points.

A release means:

The system has learned enough to be trusted.


License

This project is licensed under the MIT License.

If you distribute this software, you must keep the copyright and license information in LICENSE unchanged.

Some parts of this project may include open-source code from other sources. When a source file or folder includes its own LICENSE, that license applies to that code.


Short Definition

Skill Pilot is an AI-native Codeware system where the source code is the interface, commits are capabilities, and software evolves through use.

About

Resources

License

Contributing

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors