Meta Agent Tools
PT

Back to the catalog

skill-os

An enterprise-grade Continuous Agentic Skills Framework and operating system for AI coding agents.

Open source Open in the app JSON README (API)

About

An enterprise-grade Continuous Agentic Skills Framework and operating system for AI coding agents.

Details

Kind
Plugins
Topic
No topic detected
Publisher
naravid19
Origin
gemini
Category
ferramentas
Version
1.0.0
Stars
1
Last push
2026-05-11T16:04:21Z
Repository state
ativo
Language
Python
Added
2026-08-30 14:13:39
Updated
2026-08-30 14:13:39
Origin id
naravid19/skill-os

README

# SkillOS

SkillOS is an enterprise-grade **Continuous Agentic Skills Framework**. It acts as the definitive "operating system" for your AI coding agents, built on top of a set of composable skills and highly authoritative instructions that make sure your agent uses them.

## Quickstart

Give your agent an Operating System: [Gemini CLI](#gemini-cli), [Claude Code](#claude-code), [Cursor](#cursor), [Copilot CLI](#github-copilot-cli).

## How it works

It starts from the moment you fire up your coding agent. As soon as it sees that you're building something, it *doesn't* just jump into trying to write code. Instead, it respects the **SkillOS Bootstrapper** and hooks into the **Methodology Engine**.

SkillOS enforces a strict state machine. The agent is logically and cryptographically blocked from writing application code until the **Brainstorming** and **Planning** phases are signed off. 

It heavily emphasizes **Andrej Karpathy's LLM Coding Guidelines**:
- Think Before Coding
- Simplicity First
- Surgical Changes
- Goal-Driven Execution

Once you say "go", it leverages the **JIT Context Kernel** to dynamically load only the skills it needs (Progressive Disclosure) preventing context window bloat. The system continuously runs an `@audit_logger` telemetry to document the agent's actions into a lightweight project memory file.

If your agent runs a terminal command that fails two times consecutively, the built-in **Anti-Looping Failsafe** will force it to halt and request human intervention. Your coding agent doesn't just write code; it operates with surgical discipline.

## Installation

Installation differs by harness. The framework is designed as a portable, cross-platform extension.

### Gemini CLI

SkillOS can be installed directly as a Gemini extension:

```bash
gemini extensions install https://github.com/naravid19/skill-os
```

- Update later:
  ```bash
  gemini extensions update skill-os
  ```

### Claude Code

1. Clone the repository into your global or local tools directory:
   ```bash
   git clone https://github.com/naravid19/skill-os
   cd skill-os
   ```
2. The framework includes a native `.claude-plugin/` configuration folder. Claude Code will automatically wire up the instructions.

### Cursor

1. Clone the repository into your project directory:
   ```bash
   git clone https://github.com/naravid19/skill-os
   cd skill-os
   ```
2. The framework includes a native `.cursor-plugin/` configuration folder. Open Cursor Agent chat and the rules will automatically apply.

### GitHub Copilot CLI

SkillOS hooks natively support Copilot CLI context injection. Ensure you have the `COPILOT_CLI` environment variable configured and run the `session-start` hook.

---

## 🛠 Initialization (Required for all platforms)

After installing/cloning, you must initialize the **JIT Catalog**.
Run the indexer to scan all local and external skills and generate the `SKILL_CATALOG.json` file. The AI agent needs this file to load skills efficiently.

```bash
python scripts/indexer.py
# Or use `py scripts/indexer.py` on Windows
```

*(Optional) Configure External Skills:*
If you have a centralized folder for agent skills outside this project, set the environment variable so the indexer can find them:
- **Windows (PowerShell):** `$env:SKILLOS_EXTERNAL_DIRS="C:\path\to\your\external\skills"`
- **macOS/Linux:** `export SKILLOS_EXTERNAL_DIRS="/path/to/your/external/skills"`

## The Basic Workflow

1. **1_Brainstorming** - Activates before writing code. Refines rough ideas through questions, explores alternatives, and presents design in sections for validation.
2. **2_Planning** - Activates with approved design. Breaks work into bite-sized tasks. Every task has exact file paths, complete code, and verification steps.
3. **3_TDD_Execution** - Enforces RED-GREEN-REFACTOR. The agent must write a failing test, watch it fail, write minimal code to pass it, and commit.
4. **4_Review** - Activates when tasks complete. Verifies tests and cleans up the workspace.

**The agent checks for relevant skills before any task.** Mandatory workflows, not suggestions.

## What's Inside

### Core Architecture
- **Hooks (`hooks/`)** - Cross-platform polyglot wrappers (`run-hook.cmd`) that inject JSON context into Claude, Cursor, and Copilot on startup.
- **Scripts (`scripts/`)** - Background daemons including the `state_manager`, `indexer` (JIT Context Kernel), and `memory_manager` (Audit Telemetry).
- **Skills (`skills/`)** - The repository of authoritative commands, utilizing Progressive Disclosure.

### Skills Library
- **skill-creator** - Create new skills following the Agent Skills Framework best practices.
- **planning** - Enforces TDD and the anti-looping rules.

## Philosophy

- **Test-Driven Development** - Write tests first, always.
- **Systematic over ad-hoc** - Process over guessing.
- **Complexity reduction** - Simplicity as primary goal.
- **Evidence over claims** - Verify before declaring success.

## Contributing

The general contribution process for SkillOS is below:

1. Fork the repository
2. Switch to the `dev` branch
3. Create a branch for your work
4. Follow the `skill-creator` skill for creating and testing new and modified skills
5. Submit a PR.

## License

MIT License - see `package.json` for details.

## Community

Built and maintained by **Naravid**.

- **GitHub Repository**: https://github.com/naravid19/skill-os

More