What problem does it solve? Deciding whether a request needs a quick heuristic answer or deep multi-step analysis is inconsistent when done manually, leading to wasted tokens on trivial tasks or shallow answers on complex ones. This Skill automates that decision with a dual-process routing engine inspired by Kahneman's System 1/2 theory. ## Core Features & Use Cases - Complexity Scoring: Weighs token estimate, domain count, step count, ambiguity, and risk level into a normalized 0-1 score compared against a configurable threshold. - Escalation Rules: Automatically escalates from System 1 to System 2 on low confidence, pattern misses, high-risk keywords, multi-domain requests, or explicit flags like --think and --ultrathink. - Adaptive Learning Loop: Adjusts the routing threshold based on outcome feedback, tightening on success and loosening on failure or escalation. - Advisor Routing: Lets Sonnet-tier executors consult an Opus advisor for architecture decisions, security assessments, and complex debugging instead of full handoff. - Use Case: A developer asks Claude Code to fix a typo versus audit authentication security; the router sends the first to fast pattern matching and the second to deep deliberative analysis with sandbox verification. ## Quick Start Ask Claude Code any development question and the cognitive-router hook will automatically classify its complexity and route it to fast or deep processing.
