Edward0l1/skill-flare-discover

---
🚀 What Is SkillFetch Reloaded?
Imagine your AI coding agent isn't just a parrot—it's a blacksmith. **SkillFetch Reloaded** is the forge where your agent discovers, scores, and installs pre-forged skills from **9 distinct registries** across the internet. We don't just fetch code; we fetch *capability*. Think of it as an **App Store for agent superpowers**, where security labels replace star ratings, and pagination replaces information overload.
Born from the pragmatic vision of the original skill-fetch concept, this system evolves the single-repo fetcher into a **multi-registry skill orchestra**. Your Claude or OpenAI agent doesn't need to reinvent the wheel—it can grab a pre-tested, security-labeled skill that handles files, APIs, databases, or web scraping, all with one command.
---
✨ The Core Metaphor
> "If a normal LLM is a library of books, SkillFetch Reloaded is the librarian who can also read the books, summarize them, and hand you the exact tool you need—from any library in the world."
---
🧩 Key Features (The Superpowers)
| Feature | Description | Why It Matters |
|---|---|---|
| **Multi-Registry Discovery** | Scans 9 sources (GitHub, PyPI, npm, Hugging Face, Docker Hub, custom registries) | No vendor lock-in |
| **Security-First Labeling** | Every skill gets an AI-verified security label: Safe, Audited, Experimental, Risky | Prevents supply-chain attacks |
| **Pagination Engine** | Handles 10,000+ results gracefully with cursor-based pagination | No memory blowouts |
| **Score & Rank** | Uses weighted scoring (downloads, recency, code quality, trust) | Finds diamonds in the rough |
| **One-Command Install** | skill-fetch install [skill-id] | 2-second on-boarding |
| **Plugin Architecture** | Write your own registry plugin in 10 lines of Python | Infinite extensibility |
| **OpenAI / Claude Compatible** | Works natively with any LLM that accepts tool calls | Future-proof |
| **Responsive CLI** | Beautiful terminal UI with progress bars and colorized output | Feels premium |
| **Multilingual Support** | Docs, errors, and help in 12 languages | Global team ready |
| **24/7 Auto-Discovery Mode** | Daemon mode that watches registries for new skills | Always fresh |
---
📜 License & Legal
This project is released under the **MIT License**. You are free to use, modify, and distribute this software as you see fit, provided you include the original copyright notice.
**Copyright © 2026** — All rights reserved under the MIT License.
---
🧠 System Architecture (How It Works)
---
🖥️ Example Profile Configuration
Instead of hardcoding API keys, SkillFetch Reloaded uses a YAML profile. Here's a minimal config that works with OpenAI and Claude:
---
🔧 Example Console Invocation
Here's what a typical session looks like. The CLI is designed to feel like a conversation with a concierge:
---
💻 OS Compatibility
| Operating System | Status | Notes |
|---|---|---|
| **Linux (Ubuntu 22.04+)** | 🟢 Fully Supported | Native performance |
| **Linux (Debian 11+)** | 🟢 Fully Supported | Same as Ubuntu |
| **macOS Ventura+** | 🟢 Fully Supported | Intel & Apple Silicon |
| **macOS Monterey** | 🟡 Experimental | Some Docker features limited |
| **Windows 11** | 🟢 Fully Supported | WSL2 recommended |
| **Windows 10** | 🟡 Experimental | PowerShell 7+ required |
| **FreeBSD** | 🟠 Community Supported | No official builds yet |
| **ARM64 (Raspberry Pi)** | 🟢 Fully Supported | Slower but works |
| **Docker** | 🟢 Fully Supported | Official image available |
---
🔌 OpenAI API & Claude API Integration
SkillFetch Reloaded is designed to be **model-agnostic**. It speaks the language of both OpenAI and Claude natively.