What problem does it solve? Building a fair, well-aligned exam is hard: items must map to learning objectives, distractors must be plausible, difficulty must be controlled, and scoring must be transparent. This Skill guides an AI agent through the full exam-construction lifecycle so classroom tests function as evidence-based measurement instruments rather than loose collections of questions. ## Core Features & Use Cases - Exam and Blueprint Generation: Creates test blueprints (双向细目表), full exam papers, A/B parallel forms, answer keys, scoring rubrics, and item metadata tables aligned to course objectives and cognitive levels. - Item Review and Revision: Diagnoses MCQ and constructed-response items for distractor quality, answer uniqueness, cueing, fairness, and cognitive level, reporting issues as blocking, major, or minor. - Post-Exam Item Analysis: Computes item difficulty, corrected item-total correlation, high-low discrimination, blank rates, and distractor frequencies from response CSVs via scripts/analyze_items.py, plus a structural MCQ warning checker (scripts/validate_mcq.py). - Use Case: Given a course syllabus for Developmental Psychology, ask for a 100-point, 90-minute closed-book final exam and receive a blueprint, student-facing paper, answer key, rubrics, item metadata, and quality-review notes. ## Quick Start Use the exam-construction skill to design a 100-point, 90-minute closed-book final exam with a two-way specification table based on the attached course syllabus.
