What problem does it solve? Getting objective, structured feedback on a plan, code change, architecture decision, or UI output is hard to do consistently. This Skill runs a four-pass advisory review and returns a labeled report with a clear verdict, so weaknesses are surfaced before execution rather than after. ## Core Features & Use Cases - Four-Pass Analysis: Produces Summary, Strengths, Risks (ranked High/Medium/Low), and Fixes (must-fix vs nice-to-fix) sections. - Verdict Gating: Assigns Approved, Conditional, or Not Approved verdicts, with a rule that nothing with Medium-or-higher risk can be marked Approved. - Advisory by Design: Never blocks the user's work unless a kernel-prime violation is detected, and suggests next commands such as /plan-execute or /insight-adversarial. - Use Case: After drafting an implementation plan, run /critique to get a severity-ranked risk list and concrete fixes before handing the plan to /plan-execute. ## Quick Start Ask the agent to critique the last plan or code output by saying "review this" or running /critique.
