FDAA: File-Driven Agent Architecture v1.2.0 - Skill Verification Pipeline

Zero-Trust Verification for Agent Skills and MCP Servers This release adds the complete Skill Verification whitepaper documenting the 4-tier verification pipeline: Tier 1: Fast Pass (regex-based static analysis) Tier 2: Guard Model (LLM-as-judge semantic analysis) Tier 3: Sandbox (Docker isolated execution) Tier 4: Registry (Ed25519 signing + Merkle trees) Key contributions: W^X enforcement patterns, cryptographic signing with Ed25519, content-addressed storage with SHA-256, Merkle tree integrity verification. Part of the Fully Deterministic Agent Architecture (FDAA) framework by Substr8 Labs.