Skillware

"I know Kung Fu." - Neo

Skillware is an open-source framework and registry for modular, actionable Agent capabilities. It treats Skills as installable content, decoupling capability from intelligence.

Just as apt-get installs software and pip installs libraries, skillware installs know-how for AI agents.

---

🚀 Mission

The AI ecosystem is fragmented. Every developer re-invents tool definitions, system prompts, and safety rules for every project.

Skillware standardizes this by packaging capabilities into self-contained units that work across Gemini, Claude, GPT, and Llama.

A Skill in this framework provides everything an Agent needs to master a domain:

1. Logic: Executable Python code (the muscle).
2. Cognition: System instructions and "cognitive maps" (the mind).
3. Governance: Constitution and safety boundaries (the conscience).
4. Interface: Standardized schemas for LLM tool calling (the language).

📂 Repository Structure

This repository is organized "Corpo-Professional" style to serve as a robust foundation for enterprise-grade agents.

Skillware/
├── skillware/                  # The Core Framework Package
│   ├── core/
│   │   ├── base_skill.py       # Abstract Base Class for all skills
│   │   ├── loader.py           # Universal Skill Loader & Model Adapter
│   │   └── env.py              # Environment Management
│   └── skills/                 # The Skill Registry (Domain-driven)
│       └── finance/
│           └── wallet_screening/ # Example Complex Skill
│               ├── skill.py    # Logic
│               ├── manifest.yaml # Metadata & Constitution
│               ├── instructions.md # Cognitive Map
│               ├── card.json   # UI Presentation
│               ├── data/       # Integrated Knowledge Base
│               └── maintenance/ # Skill-Maintenance Tools
├── examples/                   # Reference Implementations
│   ├── gemini_wallet_check.py  # Google Gemini Integration
│   └── claude_wallet_check.py  # Anthropic Claude Integration
├── docs/                       # Comprehensive Documentation
│   ├── introduction.md         # Deep Dive into Philosophy
│   ├── usage/                  # Integration Guides
│   └── skills/                 # Skill Reference Cards
└── COMPARISON.md               # Vs. Anthropic Skills / MCP

⚡ Quick Start

1. Installation

Clone the repository and add it to your path (pip install coming soon).

git clone https://github.com/arpa/skillware.git
cd skillware
pip install -r requirements.txt

2. Configure Environment

Create a .env file with your keys:

ETHERSCAN_API_KEY="your_key"
GOOGLE_API_KEY="your_key"
ANTHROPIC_API_KEY="your_key"

3. "Hello World" (Gemini)

import google.generativeai as genai
from skillware.core.loader import SkillLoader
from skillware.core.env import load_env_file

# Load Env
load_env_file()

# 1. Load the Skill
# The loader reads the code, manifest, and instructions automatically
skill = SkillLoader.load_skill("finance/wallet_screening")

# 2. Configure the Agent
genai.configure(api_key=os.environ["GOOGLE_API_KEY"])
model = genai.GenerativeModel(
    'gemini-2.0-flash-exp',
    # Auto-converts manifest to Gemini Tool format
    tools=[SkillLoader.to_gemini_tool(skill)],
    # Injects the "Mind" of the skill into the Agent
    system_instruction=skill['instructions'] 
)

# 3. Execute
chat = model.start_chat(enable_automatic_function_calling=True)
response = chat.send_message("Screen wallet 0xd8dA... for risks.")
print(response.text)

📖 Documentation

• Core Logic & Philosophy: How Skillware decouples Logic, Cognition, and Governance.
• Usage Guide: Gemini: Detailed integration with Google's GenAI SDK.
• Usage Guide: Claude: Detailed integration with Anthropic's SDK.
• Skill Library: Browse available capabilities.

🤝 Contributing

We are building the "App Store" for Agents. We need professional, robust, and safe skills.

Please read CONTRIBUTING.md for our strict guidelines on folder structure, manifest schemas, and safety constitutions.

🆚 Comparison

How does this differ from the Model Context Protocol (MCP) or Anthropic's Skills repo?

• Model Agnostic: Native adapters for Gemini, Claude, and OpenAI.
• Code-First: Skills are executable Python packages, not just server specs.
• Cognitive-First: We treat instructions as a first-class citizen, equal to code.

Read the full comparison here.

---

Built by ARPA Hellenic Logical Systems