qingchunwuhui

skill-value-assessor

Evaluate whether a task is worth converting into an AI Skill using a three-dimensional assessment model (Cognitive Friction, Process Structure, Reuse Value). Use when the user wants to decide if a workflow should be automated as a Skill, asks questions like "Should I make this a Skill?", "Is this worth automating?", or explicitly invokes with /assess-skill or /skill-eval. Provides guided questioning, automatic scoring, and concrete recommendations.

qingchunwuhui 0 Updated 3mo ago

Resources

1
GitHub

Install

npx skillscat add qingchunwuhui/xianfengaiskills/skill-value-assessor

Install via the SkillsCat registry.

SKILL.md

Skill Value Assessor

Core Purpose

Help users make informed decisions about whether a manual workflow should be converted into an AI Skill by applying a structured three-dimensional assessment model.

When This Skill Is Used

This skill triggers when users:

  • Ask if a task should be made into a Skill
  • Want to evaluate automation value before investing time
  • Need guidance on prioritizing Skill development
  • Use commands like /assess-skill or /skill-eval

Assessment Workflow

Phase 1: Task Understanding

First, clearly identify the task being evaluated:

  1. Extract the task description from user input
  2. Clarify the workflow if unclear:
    • "Can you describe the steps you currently take manually?"
    • "What makes this task difficult or time-consuming?"

Phase 2: Guided Three-Dimensional Assessment

Evaluate the task across three dimensions using guided questioning. Present one dimension at a time to avoid overwhelming the user.

Dimension A: Cognitive Friction (认知摩擦力)

"How mentally draining is this task?"

Present the scoring scale:

【维度 1/3:认知摩擦力】— 人脑有多抗拒这个任务?

1分 🟢 顺手就做,不费脑子
      示例:回复"收到",简单的复制粘贴

2分 🟢 略微繁琐,但还好
      示例:发送常规邮件

3分 🟡 需要停下来想一想,或查资料
      示例:写复杂SQL,查API文档

4分 🟠 很烦,总想拖延,容易遗漏
      示例:代码审查,写技术文档

5分 🔴 极度消耗脑力,让人心累
      示例:全景盲点扫描,多语言翻译保持格式

基于你的描述 [复述任务],我初步判断可能是 [X] 分。
你同意吗?或者请告诉我你的实际感受:

Scoring logic:

  • If user provides a number (1-5), use it directly
  • If user describes feelings, map to appropriate score:
    • "easy", "simple", "quick" → 1-2
    • "annoying", "tedious", "need to think" → 3
    • "hate doing this", "always procrastinate" → 4
    • "exhausting", "error-prone", "overwhelming" → 5

Dimension B: Process Structure (结构化程度)

"How clear and repeatable is the workflow?"

Present the scoring scale:

【维度 2/3:结构化程度】— 这个流程有多清晰?

1分 🔴 完全依赖灵感,每次都不一样
      示例:写诗,画抽象画,创意设计

2分 🟠 有大致框架,但中间步骤模糊
      示例:写读后感,整理书签(无固定分类标准)

3分 🟡 有框架,部分步骤需要判断
      示例:写项目总结,代码重构

4分 🟢 步骤比较清晰,有明确的检查点
      示例:根据模板写文档,格式化数据

5分 🟢 完全标准化,有明确的 if-then 逻辑
      示例:提取摘要,格式化JSON,5W1H检查清单

关键问题:你能用"如果...那么..."的规则描述这个流程吗?
如果流程每次都差不多,且步骤明确 → 分数高
如果需要大量创意或每次都不同 → 分数低

基于 [任务描述],我认为结构化程度是 [X] 分。
你同意吗?或者描述一下你的流程:

Scoring logic:

  • Ask: "Can you describe the process as a series of if-then rules?"
  • If workflow varies significantly each time → 1-2
  • If there's a rough template but needs judgment → 3
  • If steps are mostly clear with minor variations → 4
  • If fully deterministic and can be written as checklist → 5

Critical check: If score is ≤2, proactively suggest:

⚠️ 注意:你的流程结构化程度较低([X]分)。

这可能意味着:
- 每次执行流程都需要创意或大量判断
- 难以用明确的步骤描述

建议:
• 如果可以先优化流程(建立标准化步骤),结构化程度可能提升到4-5分
• 如果流程本质上依赖创意,可能不适合做成Skill

是否需要我帮你分析如何优化流程?[Y/N]

Dimension C: Reuse Value (复用价值)

"How often will this be used, and what's the cost of errors?"

Present the scoring scale:

【维度 3/3:复用价值】— 值得折腾吗?

1分 🔴 低频低值:一年做一次,做错了也无所谓
      示例:年终总结的开头寒暄

2分 🟠 低频中值:偶尔需要,但不太重要
      示例:清理临时文件

3分 🟡 高频低值:每天都做,但很简单
      示例:整理桌面文件

4分 🟢 低频高值:不常做,但出错代价很大
      示例:服务器部署配置,合同审核

5分 🟢 高频高值:每天都做,且直接影响产出质量
      示例:代码提交规范检查,每日复盘,需求分析

请回答:
1. 这个任务多久做一次?(每天/每周/每月/偶尔)
2. 如果做错了,会有什么后果?(无所谓/有点麻烦/可能导致严重问题)

基于你的回答,我判断复用价值是 [X] 分。

Scoring logic:

  • Daily + high impact → 5
  • Daily + low impact → 3
  • Rare + high impact → 4
  • Rare + low impact → 1-2

Phase 3: Calculation & Recommendation

After collecting all three scores:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📊 评估结果汇总

任务:[任务名称]

维度评分:
  A. 认知摩擦力:[X]/5 分
  B. 结构化程度:[X]/5 分
  C. 复用价值:  [X]/5 分

总分:[XX]/15 分
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Apply decision thresholds:

Score 12-15: 立即封装 (Must Have) ✅

🎯 判定:立即封装 (Must Have)

这是 AI Skill 的甜蜜区!
• 既能极大减轻你的负担
• AI 又能执行得很好
• 投资回报率高

💡 行动建议:
1. 马上创建 Skill(使用 /skill-creator)
2. 这将成为你的核心生产力工具
3. 建议在3个月后复盘:实际使用频率是否符合预期

📚 相似案例:
[匹配相似的高分案例]

Score 8-11: 优化后封装 (Should Have) ⚠️

⚠️ 判定:优化后封装 (Should Have)

当前得分:[XX]/15

瓶颈分析:
[识别最低分的维度]

• 如果 B (结构化程度) 低:
  → 先优化流程,建立清晰的SOP
  → 手工执行2-3次验证步骤
  → 然后再考虑自动化

• 如果 C (复用价值) 低:
  → 观察1-2周,记录实际使用频率
  → 如果频率上升,重新评估

• 如果 A (认知摩擦力) 低:
  → 手动执行可能更快
  → 除非未来频率提高

💡 建议的优化路径:
[具体的优化建议]

优化后预计得分:[估算]

Score 0-7: 保持人工 (Won't Do) ❌

❌ 判定:保持人工 (Won't Do)

原因分析:
[说明为什么不适合自动化]

• 如果任务太简单(A=1-2, C=1-2)
  → 手动更快,自动化是杀鸡用牛刀

• 如果流程不清晰(B=1-2)
  → AI无法处理高度依赖创意/直觉的任务
  → 除非能先标准化流程

• 如果频率太低(C=1-2)
  → 维护Skill的成本 > 节省的时间

💡 替代方案:
[建议其他工具或方法]

Phase 4: Case Matching (Optional Enhancement)

Load references/case_library.md to find similar cases:

📚 相似案例参考:

你的任务与以下案例相似:

案例:[案例名称]
评分:[X+X+X = XX分]
判定:[结果]
关键学习:[一句话总结]

查看详情 → [[references/case_library.md#案例名称]]

Phase 5: Save Assessment (Optional)

Ask user if they want to save this assessment:

💾 是否保存此次评估记录?

保存后可以:
• 季度复盘时查看决策模式
• 追踪哪些高分Skill实际有用
• 避免重复评估相似任务

[Y] 保存到评估历史
[N] 不保存

如果选择 Y,将记录保存到评估历史文件。

If user confirms, use the template from references/assessment_template.md to create/append to the assessment history file.

Key Principles

  1. One dimension at a time - Don't overwhelm with all three dimensions at once
  2. Proactive suggestions - If you notice red flags (e.g., low structure score), suggest optimizations
  3. Concrete examples - Use examples from case library to help calibration
  4. Clear thresholds - Apply the 12/8/7 scoring thresholds consistently
  5. Context-aware - If the task description is vague, ask clarifying questions before scoring

Reference Materials

  • Detailed evaluation modelreferences/evaluation_model.md
  • Case libraryreferences/case_library.md
  • Assessment templatereferences/assessment_template.md

Load these references as needed during the assessment process.

Anti-Patterns to Avoid

❌ Don't ask all three dimensions simultaneously
❌ Don't give scores without explanation
❌ Don't skip the recommendation phase
❌ Don't ignore low structure scores (proactively suggest optimization)
❌ Don't force users to use exact numbers (accept descriptions)