What problem does it solve? When an AI assistant recommends a library, approach, or tradeoff, it tends to rubber-stamp its own reasoning on follow-up. This Skill provides a structured self-review of the immediately preceding recommendation, surfacing unstated assumptions, hidden costs, and conditions where the pick is wrong. ## Core Features & Use Cases - Scoped Recommendation Review: Locks onto the assistant's prior turn (or a user-specified pick) and restates it in one line before critiquing. - Fact Grounding: Verifies checkable repo facts and project constraints (naming policies, manual deploy steps, i18n requirements) before defending the pick. - Fresh-Eyes Subagent: Dispatches a single independent subagent with minimal context to find blind spots the self-review cannot see. - Use Case: After the assistant suggests using Wouter over React Router, the user types /why and receives a concise review covering why it matters, the real reasoning, honest gaps, alternatives, and a calibrated bottom line. ## Quick Start Type /why immediately after receiving a recommendation to get an honest pressure-test of that pick.
