What problem does it solve? AI-generated code often suffers from over-engineering, hallucinated APIs, scope creep, and superficial fixes. This Skill acts as a librarian that applies production-quality policy lenses to coding work, ensuring code meets a fixed quality bar without loading every policy document into context. ## Core Features & Use Cases - Selective policy loading: Reads a policy index and loads at most four relevant policy lenses (ai-antipattern, coding, review, existing-system-respect, design-planning) instead of all references. - Two application lenses: Applies generation lens (pre-implementation preflight and post-implementation self-check) and review lens (findings mapped to must/should/nit severities with a REJECT/Warning/APPROVE process verdict). - Default-on with proportional effort: Code work is treated as production by default; only explicit vibe/prototype/throwaway work skips it, and trivial changes get a lighter procedure without lowering the quality bar. - Use Case: When asked to implement a payment API client at production quality, the Skill reads the policy index, loads the relevant lenses, runs preflight checks (requirement match, API existence, minimal diff, contract preservation), then self-checks the result for over-implementation and unreachable code. ## Quick Start Ask the AI to implement or review code at production quality, and it will select the relevant policy lenses from the index and apply preflight and self-check before reporting.
