What problem does it solve? Single-reviewer plan audits normalize blind spots and miss critical issues. This Skill runs three independent subagent reviewers over .context/plans/*.md files, each with a distinct lens, and consolidates their divergent findings into one actionable report. ## Core Features & Use Cases - Multi-Perspective Review: Spawns Technical (feasibility, gaps), Strategic (scope, alignment, priority), and Risk (blind spots, failure modes) subagents in parallel against one shared self-contained plan brief. - Structural Validation: Checks plan frontmatter against a JSON schema and infers body-section conventions from existing plans, plus analyzes phases, tasks, and parallelization waves. - Model Routing Guidance: Detects the environment (OpenCode Zen, OpenCode Go, native Anthropic CLI, or BYOK) and asks the user which models to assign per reviewer for perspective diversity. - Finding Resolution: Classifies every finding as Editorial (applied directly) or Decision (resolved via a one-question interview recorded in a ## Decisions section, triggering adr-capture). - Use Case: Before marking a draft implementation plan ACTIVE, run the audit to catch missing rollback paths, sequencing errors, and unstated assumptions across three independent viewpoints. ## Quick Start Ask the agent to review the plan at .context/plans/<name>.md and answer the model-selection question when prompted.
