What problem does it solve? Generic code review checklists miss stack-specific bugs like Django serializers leaking columns with fields = 'all', FastAPI routes blocking the event loop, or React lists corrupted by key={index}. This Skill layers per-language and per-framework review criteria on top of a general four-domain review so findings reflect the actual idioms, ORM mechanics, and framework misconfigurations of the stack under review. ## Core Features & Use Cases - Automatic stack detection: Detects the stack by manifest files first (package.json, manage.py, go.mod, Cargo.toml, composer.json) and loads only the matching lens files, supporting React, Angular, Vue, NestJS, Python/Django/FastAPI/Celery, Go, Rust, Laravel, Java/Spring, Kotlin, Swift, .NET, C++, Ruby, Perl, ArkTS, and mobile stacks. - Ground-truth verification: Each lens ships runnable commands (ng build, gradlew, hvigorw, clang-tidy) and caps unverified findings at Medium confidence. - Noise control and false-positive discipline: Consolidates repeated findings, requires over 80% confidence before flagging, never marks pre-existing code as CRITICAL, and checks project conventions before reporting. - Use Case: When asked to review a Django pull request, the skill loads the Python and Django lenses, flags a missing select_related causing N+1 queries, verifies it against the ORM, and files the finding inside the general review's existing severity format. ## Quick Start Ask the assistant to review the code in this repository using the language-specific review lenses for the detected stack.
