What problem does it solve? Code reviews often produce vague, generic feedback like "improve error handling" that authors cannot act on. This Skill performs a rigorous, falsifiable review of code against five engineering categories and produces specific, evidence-backed findings with file:line citations, concrete triggering scenarios, and severity triage — the kind of review a senior engineer would actually write. ## Core Features & Use Cases - Five-category checklist review: Systematically checks correctness (off-by-one, race conditions, resource leaks), design & architecture (layering, coupling, abstraction fit), coding standards, robustness (boundary validation, injection, timeouts under locks), and maintainability (testability, dead code). - False-positive filtering: Cross-checks every candidate finding against a reference list of intentional simplicity, project conventions, and unreachable-state defenses so only real defects get reported. - Severity triage and structured reporting: Classifies each finding as Blocker, Should-fix, Nit, or Question, and reports it with file:line, a one-sentence defect description, and a concrete triggering scenario. - Use Case: After writing a non-trivial module — say, an async service that fetches third-party data under a shared lock — ask for a quality review. The Skill will catch issues like an unbounded await inside the lock with no timeout, cite the exact line, and explain the freeze scenario. ## Quick Start Review the module I just wrote in src/payments for correctness, design quality, and robustness issues, and report findings with severity levels.
