v2.0.0 — LLM-Powered Intelligent Verification

Added

• Enhanced Stop prompt (Cycle 3): Multi-section intelligent review covering Code Quality, Security, Research Claims, and Completeness
• Verification tag cascade (Cycle 4): Accepts multiple verification tags. Removes hard dependency on Perplexity MCP
• Optional LLM Advisory mode: Opt-in Claude Haiku analysis for deeper review (never blocks)
• Quiet mode: Minimal output when all checks pass
• Configurable Cycle 3 sections: Toggle individual review sections

Changed

• Cycle 3 prompt: context-aware security checks (test files vs production code)
• Cycle 4: tool-agnostic language for verification
• Python port mirrors all JS changes

Fixed

• Cycle 4 no longer requires Perplexity MCP specifically