What problem does it solve? Complex engineering requests often get shallow, improvised answers or over-engineered responses to simple questions. This Skill enforces a disciplined methodology that classifies every query before acting, routes it to the right depth of reasoning, and verifies results programmatically before declaring work done. ## Core Features & Use Cases - AFRP Gate: Classifies every query by type (factual, analytical, implementation, ambiguous) and audience level, then routes it to Mode 1 (inline), Mode 2 (five-stage process), or Mode 3 (multi-agent network). - Five-Stage Methodology: Context Immersion, Visionary Architecture, Ruthless Refinement, Masterful Execution with TDD, and Crystallization with lessons captured. - Integrative Merge Doctrine: Resolves PR conflicts by synthesizing both branches' intent across six merge modes instead of wholesale --ours/--theirs deletion. - Search Frugality Chain: Orders tool usage from local semantic memory (gbrain) through code-review-graph to paid web tiers, stopping at the first tier that answers. - Use Case: When asked to fix a recurring production failure, the Skill states its AFRP classification, searches local memory first, designs a minimal fix, verifies it with tests, and records a lesson for future runs. ## Quick Start Ask the agent to ultrathink a complex refactor or architecture decision and it will classify the request, plan the stages, and verify the result programmatically.
