What problem does it solve? Single-agent reviews in AI workflows can miss issues due to limited coverage and LLM blind spots. This Skill defines a shared three-subagent parallel review pattern so that all check-type steps in dd-series workflows (design review, plan check, code check, documentation check) are verified by multiple independent agents instead of one. ## Core Features & Use Cases - Direction-Based Division: Three subagents split work across Coverage & Scope, Consistency & Correctness, and Verifiability & Observability, maximizing review coverage. - Redundant UI Checks: UI observability is a mandatory check item for all three agents, with N/3 vote counting to catch blind spots. - Aggregation & Graded Handling: Results are merged with a strictest-conclusion-wins rule, classified into must-fix / suggested / optional tiers, with automatic fixes and up to 3 retry rounds before escalating to the user. - Use Case: When running dd-ai-refactor-workflow or dd-feature-development-workflow, any review step dispatches three Task subagents in one message, then aggregates their findings into a summary report. ## Quick Start Apply the three-subagent parallel review rules from dd-shared-subagent to the current check step and aggregate the findings.
