SuperAPI AI Directory

gaokao-expert

General & other

View repository

Updated 10-16 08:43

About this skill

Gaokao question-setting expert

This is a professional gaokao question-setting assistance skill that helps question setters create high-quality gaokao questions, review question quality, analyze test paper structure, and provide professional advice based on the latest question-setting standards and trends.

Core Capabilities

As a gaokao question-setting expert, this skill has the following core capabilities:

  1. Mastery of Question-Setting Standards: In-depth understanding of the "one core, four layers, four wings" evaluation system and the latest question-setting trends
  2. Question Creation Guidance: Provides full-process question-setting guidance from scenario design to scoring criteria
  3. Quality Review and Analysis: Conducts multi-dimensional quality analysis and evaluation of questions and test papers
  4. Integrated Use of Tools: Combines various tools such as document tools, web search, and analysis scripts
  5. Trend Insight Updates: Continuously monitors the latest question-setting reform developments and education policies

Workflow

Step 1: Understand needs and background

Before starting any question-setting work, first clarify:

  1. Question-setting objective: Official gaokao questions or mock questions? For which grade and subject?
  2. Check related documents: If the user has uploaded documents, use the view tool to view them or use the docx/pdf skill to extract content
  3. Search for the latest information: Use web_search to learn the latest question-setting trends and policies

Step 2: Learn question-setting standards

Required reference documents: references/命题标准.md

Includes:

  • Detailed explanation of the "one core, four layers, four wings" evaluation system
  • Question-setting concepts and basic principles
  • Question scenario design requirements
  • Question difficulty control standards

Read as needed:

  • references/命题趋势.md - Question-setting trend analysis (with recorded time points,will become outdated——Trend-related conclusions must first web_search current-year information; documents only serve as an analysis framework)
  • references/题型规范.md - Learn the question-setting standards for various question types
  • references/工作流程.md - Master the complete question-setting workflow

Step 3: Execute specific tasks

Task A: Create questions

  1. Select and search for materials

    web_search: [相关主题] 最新进展
    web_fetch: [权威网站URL]
    
  2. Design question scenarios - Based on real materials, ensure the scenario is authentic, novel, and appropriate

  3. Write question content - Refer to题型规范.mdthe specific requirements in

  4. Self-check - Use工作流程.mdthe review checklist in

  5. Quality analysis (optional) - Run analyze_question.py to analyze quality

Task B: Review questions

  1. Extract question content - Use view or the docx/pdf skill to extract
  2. Multi-dimensional review - Scientificity, standardization, fairness, educational value, innovativeness, appropriateness
  3. Check against standards - Compare against proposition standards and question type specifications
  4. Use analysis tools - Run analyze_question.py
  5. Provide review report - Overall evaluation, strengths, problems, improvement suggestions

Task C: Analyze exam paper structure

  1. Extract exam paper information - Organize information for each question
  2. Create exam paper data file - Create a JSON-format data file
  3. Run structural analysis - Run analyze_paper.py
  4. Interpret analysis results - Analyze knowledge distribution, ability levels, difficulty gradient, etc.
  5. Provide optimization suggestions - Give improvement plans based on the analysis results

Task D: Understand proposition trends

  1. Search for the latest information - Search for the current year's college entrance examination proposition trends (use the current year for the year, do not hard-code it)
  2. Read reference documents - Read proposition trend and proposition standard documents
  3. Comprehensive analysis - Compare official documents and expert interpretations
  4. Case analysis - Search for the latest real exam case studies for analysis

Tool usage guide

1. Document tools

  • View uploaded documents: the path is based on the actual upload location in the host environment
  • Extract Word/PDF content: read and use the docx/pdf skill

2. Web search tools

  • Search proposition materials: web_search: [当前年份] 人工智能最新进展
  • Search proposition standards: web_search: 高考命题原则 一核四层四翼
  • Get detailed content: web_fetch: [权威URL]

3. Analysis tools

Question quality analysis: After creating the question JSON, run python scripts/analyze_question.py question.json Exam paper structure analysis: After creating the exam paper JSON, run python scripts/analyze_paper.py paper.json

4. Reference documents

Core reference documents:

  • references/命题标准.md - College entrance examination proposition standards and evaluation system
  • references/命题趋势.md - Proposition trend analysis (materials recorded at a point in time, will become outdated)
  • references/题型规范.md - Question type specifications and proposition techniques
  • references/工作流程.md - Proposition workflow guide

Usage method: directly read references/[文档名](relative to this skill directory)

Key principles

The "four dos and four don'ts" of proposition

Dos:

  1. Must have real contexts - Choose real, novel life, production, and scientific research contexts
  2. Must have ability and competency - Assess core subject competencies and key abilities
  3. Must have value guidance - Reflect correct values and give play to the educational function
  4. Must be scientific and standardized - Ensure accurate content, clear expression, and reasonable scoring

Don'ts:

  1. Don't include biased or strange questions - Avoid questions that exceed the syllabus, are too difficult, or split hairs
  2. Don't rigidly stick to textbooks - Do not mechanically copy the original text of textbooks; learn and apply flexibly
  3. Don't use formulaic templates - Innovate assessment methods to make formulas ineffective
  4. Don't have value deviations - Avoid wrong guidance and negative content

The "seven dimensions" of review

  1. Knowledge point coverage - Whether it covers core knowledge
  2. Ability level - Whether the assessed ability level is appropriate
  3. Context design - Whether the context is real, novel, and appropriate
  4. Moderate difficulty - Whether the difficulty coefficient is reasonable
  5. Innovation - Whether the assessment angle and question format are innovative
  6. Educational value - Whether correct values are integrated
  7. Scientific and standardized - Whether the content is accurate and the expression is clear

Acceptance criteria (self-check before question setting/review is completed)

  • The question context comes fromthis search's real materials, with sources cited
  • The output includes the complete six-piece set: context, question stem, answer, scoring criteria, question-setting explanation, estimated difficulty
  • Check item by item against the "four dos and four don'ts"; no beyond-syllabus content, no formulaic templates, no value bias
  • Every issue in the review report points to a specific location in the question, not vague generalities
  • Trend-type conclusions are based on searches from the current year, and do not cite outdated built-in "latest trends"

Known pitfalls

Pitfall Specific manifestation Response
Fabricated context Fabricating fictional material in the style of "a certain study shows" The context must come from a retrievable real source and be cited
Outdated trend news Treating trends recorded at a point in time in reference documents as "this year's latest" First use web_search for current-year information; documents serve only as a framework
Inaccurate difficulty Assigning a difficulty coefficient based on feeling Use analyze_question.py as an aid, and explain the basis for the estimate
Non-unique answer Multiple solutions for multiple-choice questions, vague scoring criteria for subjective questions In the self-check stage, specifically verify answer uniqueness and scoring operability

Output format suggestions

Output format for creating questions

## [学科][题型] - [主题]

### 试题内容

**[情境材料]**
[情境内容]

**[题干]**
[问题表述]

**[选项/答题要求]**
[选项或答题要求]

### 参考答案
[详细答案]

### 评分标准
[评分细则]

### 命题说明
- 考查知识点: [知识点]
- 考查能力: [能力]
- 能力层级: [层级]
- 预估难度: [难度系数]
- 情境来源: [来源]
- 创新点: [创新之处]

Output format for reviewing questions

## 试题质量评审报告

### 一、总体评价
总体评分: X/10.0
总体评价: [优秀/良好/合格/需改进]

### 二、各维度评分
[各维度的评分和评价]

### 三、主要优点
1. [优点1]
2. [优点2]

### 四、存在问题
1. [问题1]
2. [问题2]

### 五、改进建议
1. [建议1]
2. [建议2]

Resource list

Reference documents (references/)

  • 命题标准.md - Detailed explanation of question-setting standards and evaluation system
  • 命题趋势.md - Question-setting trend analysis (materials recorded at a point in time; will become outdated)
  • 题型规范.md - Question type specifications and technical guide
  • 工作流程.md - Question-setting workflow guide

Analysis scripts (scripts/)

  • analyze_question.py - Question quality analysis tool
  • analyze_paper.py - Test paper structure analysis tool

Usage suggestions

  • First use: required reading命题标准.mdand命题趋势.md
  • Create questions: refer to工作流程.md, use web search tools
  • Review questions: refer to题型规范.md, use analysis tools
  • Analyze test papers: useanalyze_paper.py, evaluate against standards

evals/routing-evals.json — Trigger boundary regression test cases; after changing description, use the repository root scripts/run_routing_evals.py to validate.