Agent Skills
› cosmicstack-labs/mercury-agent-skills
› assessment-design
assessment-design
GitHub指导设计公平有效的评估方案，涵盖诊断、形成性、总结性及真实性评估类型。提供分析性与整体量规设计最佳实践，强调明确标准与示例。阐述及时、具体、可操作的反馈原则，并介绍基于能力的评估模式，支持个性化学习路径与掌握导向评价。
Trigger Scenarios
设计课堂测验或考试
创建评分量规
撰写学生反馈
规划项目式评估
Install
npx skills add cosmicstack-labs/mercury-agent-skills --skill assessment-design -g -y
SKILL.md
Frontmatter
{
"name": "assessment-design",
"metadata": {
"tags": [
"assessment",
"education",
"evaluation",
"rubrics",
"feedback"
],
"author": "cosmicstack-labs",
"version": "1.0.0",
"category": "education-learning"
},
"description": "Formative vs summative, rubric design, authentic assessment, competency-based evaluation, and feedback"
}
Assessment Design
Create fair, effective assessments that measure real learning.
Assessment Types
| Type | Purpose | Frequency | Examples |
|---|---|---|---|
| Diagnostic | Find starting point | Before instruction | Pre-test, KWL chart |
| Formative | Guide learning | Ongoing | Exit tickets, quizzes, discussion |
| Summative | Evaluate achievement | End of unit | Exam, project, presentation |
| Authentic | Real-world application | Culminating | Portfolio, case study, simulation |
| Competency-based | Demonstrate mastery | On demand | Performance tasks, certifications |
Rubric Design
Analytic Rubric (Preferred)
| Criteria | 1 - Novice | 2 - Developing | 3 - Proficient | 4 - Exemplary |
|---|---|---|---|---|
| Clarity | Unclear, confusing | Somewhat clear | Clear and organized | Exceptionally clear |
| Evidence | No support | Limited support | Relevant support | Compelling, diverse evidence |
| Analysis | Summary only | Some analysis | Good analysis | Deep, nuanced insight |
Holistic Rubric
Single score based on overall quality. Faster but less specific.
Best Practices
- Define 3-5 criteria aligned to learning objectives
- Use observable, specific language (not "good" or "poor")
- Include examples or exemplars for each level
- Share rubrics before the assessment
- Calibrate with other assessors before grading
Feedback Principles
- Timely: Within 24-48 hours or before next similar task
- Specific: "Your thesis statement clearly argues X" (not "good job")
- Actionable: "Try using a clearer topic sentence at paragraph start"
- Kind: Address the work, not the person
- Forward-looking: "For next time, consider..."
Competency-Based Assessment
- Learners advance upon demonstration of mastery
- Multiple assessment opportunities
- No penalty for first attempts
- Focus on skills and application, not time spent
- Personalized learning paths
Version History
- 38e2523 Current 2026-07-05 19:39
