What problem does it solve? It provides a single, unified pre-commit review gate that verifies whether code changes correctly implement the task's PRD, design, and implementation plans, whether underlying assumptions about APIs, components, and data hold, and whether the change is complete and spec-compliant—without mixing up root-cause classification or severity. ## Core Features & Use Cases - Depth Routing: Automatically selects light or full review depth based on requested intent, hard-full signals (API/schema/security changes), and whether the impact surface is closed, with one-way escalation from light to full. - Three-Dimension Audit: Checks planning-artifact implementation, implementation assumptions (API contracts, component context, data history, data flow), and completeness/spec compliance with shared verification evidence. - Classified Findings Model: Separates main-path issues (CHK-) from fallback/defensive-path issues (FBK-) by root cause before assigning P0/P1/P2 severity, and supports a whitelisted DOC-* channel for auto-remediating low-risk factual documentation drift. - Use Case: Before committing a feature branch, ask for a check-all; the skill reads the task's prd.md/design.md/implement.md, traces the diff against them, runs targeted verification, and reports CHK/FBK findings with evidence, severity, and a single disposition prompt. ## Quick Start Ask the assistant to run a check-all review on the current workspace changes before committing, optionally requesting a light or full check.
