Back to Details

k2f-exam-paper

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

K2F Exam Paper — AI-Native Aesthetic Design

K2F 试卷——AI 原生美学设计

Purpose

目的

Teach agents to blueprint an exam paper before layout: syllabus-faithful scope, stable format fingerprints, fair point weighting, and prompts students can parse under clock pressure—not study sheets, predicted questions, or item banks pasted without editorial structure.
教授 Agent 在排版前对试卷进行蓝图规划：严格遵循教学大纲范围、采用稳定的格式指纹、公平分配分值权重，并确保提示学生在时间压力下能够准确理解。它不是学习指南、预测题或未经编辑设计的题库。

Built on K2F

基于 K2F

This skill produces exam papers as 
.K2F
packages. Follow the k2f skill for authoring: unpack or init a workspace, edit the semantic tree, run pack/verify, then export. K2F keeps agents on meaning (sections, items, stimuli, response roles)—not pixel coordinates—while the engine compiles layout, typography, and theme.
Before you build, install the core skill: 
npx skills add suzheng/k2f --skill k2f
.
此 skill 生成 
.K2F
包。请遵循 k2f skill 进行创作：解包或初始化工作区，编辑语义树，运行 pack/verify，然后导出。K2F 让 Agent 专注于含义（章节、题目、材料、作答角色），而不是像素坐标；引擎负责编译版面、排版和主题。
开始构建前，请安装核心 skill：
npx skills add suzheng/k2f --skill k2f
。

When to Use

何时使用

  • An instructor or instructional designer needs a deliverable exam document (in-class, take-home, or hybrid).
  • Inputs include syllabus topics, mark weightings, learning outcomes, approved readings, or past exams to match style—not when the user only wants study notes or “what might be on the exam.”
  • Distinct from k2f-worksheet (low-stakes practice) and k2f-lesson-plan (teaching sequence).
  • 教师或教学设计师需要一份可交付的试卷文档（闭卷、带回家完成或混合形式）。
  • 输入内容包括教学大纲主题、分值权重、学习成果、指定的阅读材料，或用于匹配风格的往年试卷；如果用户只想要学习笔记或“可能考什么”，则不应使用。
  • 不同于 k2f-worksheet（低风险练习）和 k2f-lesson-plan（教学顺序安排）。

Workflow

工作流

Step 1. Generate the content_and_design.md

第 1 步：生成 content_and_design.md

One blueprint pairing assessment content with aesthetic intent. Register shared facts once; every question references stable keys so stimuli and numbering cannot drift.
Sub-steps
  1. Exam lock — Course, assessment title, sitting length, total points, open/closed/take-home rules, and allowed materials. If past papers exist, note format fingerprints (question count, style mix, fact-pattern density, call style) to match or deliberately change—state which.
  2. Coverage map — Table syllabus topics → target points (%). Prefer explicit syllabus weightings; otherwise infer from contact hours and learning outcomes. Flag topics taught but under-weighted on purpose vs accidentally omitted.
  3. Information architecture — Layer top to bottom: identification; instructions once (time, materials, integrity, labeling); optional sections when styles differ; questions with stable IDs and points that sum to the exam lock; stimuli in shared content when multiple sub-parts share facts; response contract per sub-part (type + space budget); scoring outline in a separate role (model answers, keywords, partial credit)—not on the student layer unless requested.
  4. Item design — One assessable outcome per sub-part; split compound calls. Match question style to the map: recall, application, issue-spotting, policy, quantitative, MCQ. Order within a section: warm-up → anchor items → discriminating items unless spiral review is intended. When source readings bound scope, every item must be answerable from syllabus + named materials; mark 
    [OUT OF SCOPE]
    gaps instead of inventing content.
  5. Pattern fidelity — When matching past papers, note stable vs variable elements (topic emphasis, style mix, fact density). Preserve structure, never verbatim prior text.
  6. Design intent — Tone: formal, neutral; no coaching fluff. Hierarchy: question numbers and points scan first; stimuli grouped under shared headings. Density: tight stems, generous handwriting space. Consistency: parallel verbs and labeling. Palette: print-safe B&W; one accent max for section warnings.
Rules & constraints
  • Syllabus is the ceiling — Do not test untaught topics unless the user explicitly expands scope.
  • Points must sum — Section totals, question totals, and exam lock agree; show the arithmetic in the blueprint.
  • Single instructions block — Per-question repeats of “show your work” belong in the block once unless a section truly differs.
  • Honest difficulty — If time or page budget is tight, drop items or narrow coverage; do not shrink answer space below what the call requires.
  • Original items — Source material informs scope; never lift publisher or prior-year text verbatim without rights.
  • Separate roles — Student paper vs scoring outline vs optional cover sheet are distinct artifacts in the design doc.
Common failure modes
  • Unstated facts required → extend shared registry or cut the item.
  • Compound questions → split sub-parts and re-weight points.
  • Rubric on student layer → move to scoring outline.
  • Weighting drift vs coverage map → rebalance or revise the map.
  • Vague “Discuss” calls → add length, format, and evidence expectations.
  • Format mismatch (take-home vs timed) → fix exam lock and instructions.
Quality checklist
  • Exam lock complete; points sum to total; materials policy clear.
  • Coverage map aligned with every question; orphan topics cut or justified.
  • Shared registry lists all cross-item facts, data, and stimuli with keys.
  • Each sub-part has points, response type, and adequate space.
  • Instructions appear once; numbering stable for scoring outline cross-reference.
  • Scoring outline keyed by question ID; keywords/partial credit where constructed responses apply.
  • Design intent covers tone, hierarchy, density, consistency, palette.
  • Original items only; out-of-scope gaps flagged.
创建一份将考核内容与美学意图配对的蓝图。共享事实只注册一次；每道题引用稳定的键，确保材料与题号不会发生漂移。
子步骤
  1. 考试设定——确定课程、考核名称、考试时长、总分、开卷/闭卷/带回家完成规则以及允许使用的材料。如果存在往年试卷，请记录其格式指纹（题量、题型组合、事实模式密度、作答指令风格），以便匹配或有意改变这些特征，并明确说明采用哪种方式。
  2. 覆盖映射——用表格呈现教学大纲主题与目标分值（%）的对应关系。优先采用教学大纲中明确给出的权重；否则根据课时和学习成果进行推断。对于已讲授但有意赋予较低权重的主题与意外遗漏的主题，应分别标明。
  3. 信息架构——从上到下分层组织内容：身份识别信息；一次性说明（时间、材料、诚信要求、标注方式）；题型不同时可加入可选的章节；带有稳定 ID 且分值之和等于考试总分的题目；多个子题共享事实时将材料放入共享内容；为每个子题设置作答约定（类型与篇幅预算）；将评分大纲放在独立角色中（参考答案、关键词、部分分），除非用户另有要求，否则不要出现在学生层。
  4. 题目设计——每个子题只考查一个可评价的学习成果；拆分包含多个任务的要求。根据覆盖映射匹配题目风格：记忆、应用、问题识别、政策、定量、MCQ。章节内的排序应为：热身题 → 核心题 → 区分度题，除非有意采用螺旋复习。当指定阅读材料界定了考查范围时，每道题都必须能够依据教学大纲和指定材料作答；对于材料不足之处，应标记 
    [OUT OF SCOPE]
    ，而不是自行补充内容。
  5. 模式一致性——匹配往年试卷时，记录稳定要素与可变要素（主题重点、题型组合、事实密度）。应保留其结构，但绝不逐字复用往年文本。
  6. 设计意图——语气：正式、中性；不加入教学引导式废话。层级：优先突出题号和分值，便于快速浏览；共享材料按统一标题归组。密度：题干紧凑，并为手写作答留出充足空间。一致性：使用平行的动词和标注方式。配色：采用适合打印的黑白方案；最多使用一种强调色标示章节警示。
规则与约束
  • 教学大纲是上限——除非用户明确扩展考查范围，否则不要考查未讲授的主题。
  • 分值必须一致——章节总分、题目总分和考试设定中的总分必须一致，并在蓝图中展示计算过程。
  • 仅使用一个说明区块——除非某个章节确有不同要求，否则每道题重复出现的“展示过程”等说明应只在区块中写一次。
  • 如实控制难度——如果时间或页数预算有限，应减少题目或缩小考查范围；不要将作答空间压缩到不足以完成要求。
  • 原创题目——来源材料可用于界定考查范围；未经授权，不得逐字使用出版商或往年试题中的文本。
  • 区分角色——学生试卷、评分大纲和可选封面页应在设计文档中作为不同产物分别处理。
常见失败模式
  • 题目依赖未说明的事实 → 扩展共享事实注册表或删除该题。
  • 题目包含多个任务 → 拆分子题并重新分配分值。
  • 学生层出现评分标准 → 将其移至评分大纲。
  • 分值权重与覆盖映射不一致 → 重新平衡或修改映射。
  • 作答要求过于笼统，如“讨论” → 补充篇幅、格式和证据要求。
  • 形式不匹配（带回家完成与限时考试） → 修正考试设定和说明。
质量检查清单
  • 考试设定完整；分值之和等于总分；材料政策清晰。
  • 覆盖映射与每道题保持一致；无对应题目的主题已删除或有明确理由。
  • 共享事实注册表通过稳定的键列出所有跨题事实、数据和材料。
  • 每个子题均标注分值、作答类型和充足的作答空间。
  • 说明仅出现一次；编号稳定，可供评分大纲交叉引用。
  • 评分大纲按题目 ID 建立对应关系；建构性作答题包含关键词和部分分规则。
  • 设计意图涵盖语气、层级、密度、一致性和配色。
  • 仅包含原创题目；已标记超出范围的内容缺口。

Step 2. Use the content_and_design.md and k2f skill to generate a K2F file and export to an appropriate file format.

第 2 步：使用 content_and_design.md 和 k2f skill 生成 K2F 文件并导出为适当的文件格式

Follow k2f to instantiate the exam from the blueprint, apply student vs instructor roles from the design doc, verify, and export print-ready PDF or other deliverables.
遵循 k2f 的流程，根据蓝图实例化试卷，应用设计文档中定义的学生版与教师版角色，执行验证，并导出为可打印的 PDF 或其他适当的交付格式。