Back to Details

skill-reviewer

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Skill Reviewer

Skill 审查工具

Review Skill definition quality against best practices.
对照最佳实践审核Skill定义质量。

Default Mode: Definition Review

默认模式：定义审核

CRITICAL: When user says "review skill", ALWAYS default to Definition Review mode. Only switch to Execution Review when user explicitly mentions execution-related keywords.
Default behavior: Check the Skill definition quality (SKILL.md, structure, content).
Execution Review: Only when user explicitly mentions:
  • "analyze execution"
  • "review trace"
  • "evaluate execution process"
  • "执行过程评估"
ModeWhen to UseInput
Definition Review (Default)"review skill", "check quality", "validate"Skill folder path
Execution Review (Explicit)"analyze execution", "review trace"Execution trace + goal
重要提示：当用户说「review skill」时，始终默认使用定义审核模式。仅当用户明确提及执行相关关键词时，才可切换到执行审核模式。
默认行为：检查Skill定义质量（SKILL.md、结构、内容）。
执行审核：仅当用户明确提及以下内容时启用：
  • "analyze execution"
  • "review trace"
  • "evaluate execution process"
  • "执行过程评估"
模式适用场景输入
定义审核（默认）"review skill", "check quality", "validate"Skill文件夹路径
执行审核（需明确指定）"analyze execution", "review trace"执行轨迹 + 目标

Example

示例

User: "Review my skill at ./my-skill/"
Actions:
  1. Read folder structure: 
    ls -la ./my-skill/
  2. Check SKILL.md exists and parse YAML frontmatter
  3. Run 20-item checklist (S1-S4, F1-F5, C1-C8, T1-T3)
  4. Generate review report
Result:
undefined
用户: "Review my skill at ./my-skill/"
执行动作:
  1. 读取文件夹结构: 
    ls -la ./my-skill/
  2. 检查SKILL.md是否存在并解析YAML frontmatter
  3. 运行20项检查清单（S1-S4、F1-F5、C1-C8、T1-T3）
  4. 生成审核报告
结果:
undefined

🔍 Skill Review: my-skill

🔍 Skill Review: my-skill

Category✅ Pass⚠️ Warn❌ Fail
Structure400
Format500
Content620
Trigger300
Total1820

---
Category✅ Pass⚠️ Warn❌ Fail
Structure400
Format500
Content620
Trigger300
Total1820

---

Mode A: Definition Review

模式A：定义审核

Purpose: Check if a Skill follows best practices and standards.
用途：检查Skill是否符合最佳实践和规范。

When to Use

适用场景

  • Before publishing a new Skill
  • After modifying a Skill
  • Want to improve Skill quality
  • 发布新Skill之前
  • 修改Skill之后
  • 想要提升Skill质量

Input

输入要求

Required: Skill folder path (e.g., ./skill/ or ./my-skill/)
Required: Skill folder path (e.g., ./skill/ or ./my-skill/)

Four-Category Checklist

四类检查清单

CategoryFocusItems
📁 StructureFile/folder layoutS1-S4
📋 FormatYAML frontmatterF1-F5
📝 ContentInstructions qualityC1-C8
🎯 TriggerActivation designT1-T3
See 
references/definition-checklist.md
for complete 20-item checklist
分类检查重点检查项
📁 结构文件/文件夹布局S1-S4
📋 格式YAML frontmatterF1-F5
📝 内容说明质量C1-C8
🎯 触发方式激活逻辑设计T1-T3
完整20项检查清单请查看
references/definition-checklist.md

Workflow

工作流程

Step 1: Read Skill folder structure
        ↓
Step 2: Check Structure (S1-S4)
        ↓
Step 3: Check Format (F1-F5)
        ↓
Step 4: Check Content (C1-C8)
        ↓
Step 5: Check Trigger (T1-T3)
        ↓
Step 6: Output report
Step 1: Read Skill folder structure
        ↓
Step 2: Check Structure (S1-S4)
        ↓
Step 3: Check Format (F1-F5)
        ↓
Step 4: Check Content (C1-C8)
        ↓
Step 5: Check Trigger (T1-T3)
        ↓
Step 6: Output report

Output Format

输出格式

markdown
undefined
markdown
undefined

🔍 Skill Review: {skill-name}

🔍 Skill Review: {skill-name}

Path: 
{skill-path}
Path: 
{skill-path}

📊 Summary

📊 Summary

Category✅ Pass⚠️ Warn❌ Fail
Structurexxx
Formatxxx
Contentxxx
Triggerxxx
Totalxxx
Category✅ Pass⚠️ Warn❌ Fail
Structurexxx
Formatxxx
Contentxxx
Triggerxxx
Totalxxx

📁 Structure

📁 Structure

  • ✅ S1 SKILL.md exists — Description
  • ⚠️ S2 Item — Description
    💡 Suggestion
(Continue for all categories...)
  • ✅ S1 SKILL.md exists — Description
  • ⚠️ S2 Item — Description
    💡 Suggestion
(Continue for all categories...)

💡 Review Comments

💡 Review Comments

Strengths

Strengths

  • Point 1
  • Point 2
  • Point 1
  • Point 2

Improvements

Improvements

  • Point 1
  • Point 2
  • Point 1
  • Point 2

Best Practice Alignment

Best Practice Alignment

  • Point 1

> See `references/definition-report.md` for complete template

---
  • Point 1

> 完整模板请查看`references/definition-report.md`

---

Mode B: Execution Review

模式B：执行审核

Purpose: Analyze Skill/Tool/Agent execution through three-layer progressive analysis.
L1 Engineering Correctness ──▶ L2 Goal Achievement ──▶ L3 Optimization Space
   Is execution correct?         Is effect good?          Can it be better?
用途：通过三层递进分析，对Skill/Tool/Agent的执行情况进行分析。
L1 Engineering Correctness ──▶ L2 Goal Achievement ──▶ L3 Optimization Space
   Is execution correct?         Is effect good?          Can it be better?

When to Use

适用场景

  • Skill/Tool execution results don't meet expectations
  • Agent behavior is abnormal and needs troubleshooting
  • Want to systematically improve implementation
  • Skill/Tool执行结果不符合预期
  • Agent行为异常需要排查
  • 想要系统性优化实现逻辑

Four Input Dimensions

四个输入维度

InputDescriptionRequired
Execution TraceTool calls, output, errors✅ Yes
Execution GoalOriginal prompt✅ Yes
Implementation ReferenceSkill/Tool definitionOptional
Agent ImplementationAgent sourceOptional
See 
references/input-guide.md
for details
输入项说明是否必填
执行轨迹工具调用、输出、错误信息✅ 是
执行目标原始prompt✅ 是
实现参考Skill/Tool定义可选
Agent实现Agent源码可选
详情请查看
references/input-guide.md

Three-Layer Analysis

三层分析逻辑

L1: Engineering Correctness

L1：工程正确性

Question: Is execution correct?
  • Are tool calls successful?
  • Do scripts execute without errors?
  • Any Linter/syntax errors?
  • Any infinite loops?
Judgment: If errors found → Report and stop, fix L1 first
核心问题：执行过程是否正确？
  • 工具调用是否成功？
  • 脚本执行是否无报错？
  • 是否存在Linter/语法错误？
  • 是否存在无限循环？
判定规则：如果发现错误→上报问题并终止分析，先修复L1层面问题

L2: Goal Achievement

L2：目标达成度

Question: Is the effect good?
  • Is the goal correctly understood?
  • Are key steps executed?
  • How is output format/quality?
Rating: ✓✓✓ Excellent / ✓✓ Good / ✓ Pass / ✗ Fail
核心问题：执行效果是否达标？
  • 目标是否被正确理解？
  • 关键步骤是否都已执行？
  • 输出格式/质量如何？
评级标准：✓✓✓ 优秀 / ✓✓ 良好 / ✓ 合格 / ✗ 不合格

L3: Optimization Space

L3：优化空间

Question: Can it be better? (Prerequisite: L1 pass + L2 at least pass)
[MUST] Check these dimensions:
  • Efficiency: Redundant calls? Token waste? Repeated reads?
  • Implementation: Better tools? Clearer description? Scripts simplifiable?
  • Conciseness: Redundant content? Over-explanation?
See 
references/analysis-dimensions.md
for details
核心问题：是否可以做得更好？（前提：L1通过 + L2至少达到合格）
[必须]检查以下维度：
  • 效率：是否存在冗余调用？Token浪费？重复读取？
  • 实现：是否有更合适的工具？说明是否更清晰？脚本是否可简化？
  • 简洁性：是否存在冗余内容？过度解释？
详情请查看
references/analysis-dimensions.md

Workflow

工作流程

Step 1: Collect inputs
        ↓
Step 2: Read implementation reference (if available)
        ↓
Step 3: L1 scan → Report if issues found
        ↓
Step 4: L2 comparison → Evaluate achievement
        ↓
Step 5: L3 optimization → Generate suggestions
        ↓
Step 6: Output report
        ↓
Step 7: (Optional) Generate execution plan if user adopts suggestions
See 
references/execution-guide.md
for execution plan generation
Step 1: Collect inputs
        ↓
Step 2: Read implementation reference (if available)
        ↓
Step 3: L1 scan → Report if issues found
        ↓
Step 4: L2 comparison → Evaluate achievement
        ↓
Step 5: L3 optimization → Generate suggestions
        ↓
Step 6: Output report
        ↓
Step 7: (Optional) Generate execution plan if user adopts suggestions
执行计划生成规则请查看
references/execution-guide.md

Output Format

输出格式

markdown
undefined
markdown
undefined

Execution Review Report

Execution Review Report

Basic Information

Basic Information

ItemContent
Analysis Target{Skill/Tool/Agent}
Execution Goal"{prompt}"
ItemContent
Analysis Target{Skill/Tool/Agent}
Execution Goal"{prompt}"

L1: Engineering Correctness ✓/✗

L1: Engineering Correctness ✓/✗

[Check result table]
[Check result table]

L2: Goal Achievement ✓✓

L2: Goal Achievement ✓✓

[Achievement status + Rating]
[Achievement status + Rating]

L3: Optimization Space 💡

L3: Optimization Space 💡

[Suggestions by dimension]
[Suggestions by dimension]

Summary

Summary

[Core issues + Priority improvements]

> See `references/report-templates.md` for complete template

---
[Core issues + Priority improvements]

> 完整模板请查看`references/report-templates.md`

---

Troubleshooting

问题排查

Error: "SKILL.md not found"

错误："SKILL.md not found"

Cause: File missing or named incorrectly (SKILL.MD, skill.md, etc.) Solution: Create 
SKILL.md
(case-sensitive) in skill folder root
原因：文件缺失或命名错误（如SKILL.MD、skill.md等） 解决方案：在Skill文件夹根目录创建大小写正确的
SKILL.md
文件

Error: "YAML parse failed"

错误："YAML parse failed"

Cause: Invalid YAML syntax in frontmatter Solution: Check for:
  • Missing 
    ---
    delimiters
  • Incorrect indentation
  • Unescaped special characters in description
原因：frontmatter中的YAML语法无效 解决方案：检查以下内容：
  • 缺少
    ---
    分隔符
  • 缩进错误
  • 描述中存在未转义的特殊字符

Error: "name field invalid"

错误："name field invalid"

Cause: name contains spaces, capitals, or special characters Solution: Use kebab-case (e.g., 
my-skill-name
)
原因：name包含空格、大写字母或特殊字符 解决方案：使用kebab-case命名（例如：
my-skill-name
）

Warning: "SKILL.md too long"

警告："SKILL.md too long"

Cause: File exceeds 300 lines Solution: Move detailed content to 
references/
folder
原因：文件超过300行 解决方案：将详细内容移动到
references/
文件夹

References

参考文档

DocumentContent
references/definition-checklist.md
Definition review: 20-item checklist
references/definition-report.md
Definition review: report template
references/input-guide.md
Execution review: four input dimensions
references/analysis-dimensions.md
Execution review: three-layer checklist
references/scenarios.md
Typical scenario guide
references/report-templates.md
Execution review: report templates
references/execution-guide.md
Execution plan generation guide
文档路径内容说明
references/definition-checklist.md
定义审核：20项检查清单
references/definition-report.md
定义审核：报告模板
references/input-guide.md
执行审核：四个输入维度说明
references/analysis-dimensions.md
执行审核：三层检查清单
references/scenarios.md
典型场景指南
references/report-templates.md
执行审核：报告模板
references/execution-guide.md
执行计划生成指南