What problem does it solve? When AI assists with unfamiliar technical work, users can lose the understanding needed to review, verify, and own decisions about the output. This Skill calibrates AI support so work keeps moving while the user retains or recovers task-specific understanding, without turning every task into a lesson or blocking clear authorized work. ## Core Features & Use Cases - Iterative calibration cycle: Wraps the originating workflow with an 8-step cycle that identifies decision-relevant understanding gaps, selects a support method, and returns to the task once calibration no longer changes the next action. - Decision ownership separation: Keeps objectives, scope, risk tolerance, authorization, and final adoption with the user while AI performs evidence gathering, analysis, and authorized work. - Focused understanding checkpoints: Uses targeted explanation, choice, or verification prompts only when the answer could change the next action, never fixed quizzes or study plans. - Use Case: A user must approve an AI-generated concurrency fix but cannot explain the lock boundary. The Skill explains the mutex and callback reasoning, identifies the evidence needed for approval, and leaves the final adoption decision with the user. ## Quick Start Ask the AI to help you understand the risky parts of an AI-generated change while continuing the review, for example: help me understand why this fix works and what evidence I need before approving it.
