proveo/skills

Immutable, Decentralized Context Strategies for AI.

This project implements a "Proof of Skill" architecture. It moves context management (prompt templates, strategies, heuristics) from local files to a decentralized registry, ensuring transparency, provenance, and censorship resistance.

Architecture

The system operates in three distinct layers to balance Trust (Web3) with Performance (Web2).

Layer 1: The Store (Web3)

• Storage: IPFS (Content-Addressable Storage). Holds the actual strategy files (JSON, Text, Python).
• Registry: EVM Smart Contract (L2). Maps a Skill_ID to an IPFS_CID and Author.
• Role: The immutable "Source of Truth."

Layer 2: The Indexer (Web2)

• Watcher: A service that listens to blockchain events.
• Cache: A SQL/Vector database that indexes strategies for fast retrieval.
• Role: High-performance caching and searchability.

Layer 3: The Bridge (Access)

• API: A REST/GraphQL API for AI Agents to query context.
• MCP: A Model Context Protocol adapter for local IDE integration.
• Role: Low-latency delivery to the LLM.

Usage

1. Publish: Authors upload strategies via CLI.
2. Index: The system automatically detects and indexes the update.
3. Consume: AI Agents query the Bridge API to retrieve the best context for the current task.