What problem does it solve? Engineering decisions often inherit assumptions, legacy constraints, and industry conventions that were never verified against the actual outcome required. This Skill forces architecture, debugging, migration, and technology-selection decisions to be rebuilt from observable facts and binding invariants, exposing disguised assumptions before they become costly commitments. ## Core Features & Use Cases - Evidence classification: Labels every premise as FACT, INVARIANT, ASSUMPTION, UNKNOWN, CONVENTION, or PREFERENCE, with strict tests before anything earns invariant status. - Assumption challenging: Explicitly interrogates backward compatibility, existing architecture, change size, reversibility, vendor defaults, and sunk-cost work. - Reconstruction and falsification: Builds the smallest sufficient system from verified premises, includes the zero option, and runs counterexamples and pre-mortems before execution. - Use Case: When asked to add a caching layer to fix slow queries, the Skill first verifies whether the queries, schema, or data model are the actual root cause, and may conclude the cache should never be built. ## Quick Start Ask the AI to use first-principles thinking to challenge the current design of your system and rebuild the solution from verified fundamentals.