What problem does it solve? Standard code reviews check whether code works, but never ask whether the code should exist, whether the architecture is right, or whether the implementation fits the design. This Skill fills that gap by arguing about design intent rather than hunting bugs. ## Core Features & Use Cases - Three-lens design critique: Evaluates every finding across what was built (right artifact?), architecture (right shape?), and implementation (right code inside?), with explicit verdicts like don't-build, delete, rewrite, or wrong-but-keep. - Full-coverage codebase sweep: Partitions the repository into regions, reads every region via fresh-context subagents, and uses git history only to rank findings rather than to select what gets read. - Argumentative dialogue, not a report: Leads with the strongest claim, steelmans the existing design before attacking it, concedes to new facts, and refuses to concede to pushback alone. - Use Case: Before building a new caching layer, run the Skill on the spec — it may ladder the want from "caching layer" down to "one slow query" and save 400 lines of code. ## Quick Start Ask the AI to run lamdan on the whole repository to argue about whether the current design is the right thing to have built.
