Back to Details

skill-auditor

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Skill Auditor

Skill 审核器

Comprehensive assessment of your skill ecosystem with actionable recommendations.
为你的Skill生态系统提供全面评估及可落地改进建议。

Workflow

工作流

Phase 1: Inventory

第一阶段：清单整理

Gather all skills from:
  1. Global skills: 
    ~/.claude/skills/*/SKILL.md
  2. Project skills: 
    ./skills/*/SKILL.md
    or 
    ./.claude/skills/*/SKILL.md
For each skill, extract:
  • Name and description from frontmatter
  • Line count of SKILL.md
  • Presence of reference files, scripts, assets
  • Any non-standard frontmatter fields
从以下位置收集所有Skill：
  1. 全局Skill：
    ~/.claude/skills/*/SKILL.md
  2. 项目级Skill：
    ./skills/*/SKILL.md
    或 
    ./.claude/skills/*/SKILL.md
针对每个Skill，提取以下信息：
  • 来自frontmatter的名称和描述
  • SKILL.md的行数
  • 参考文件、脚本、资源的存在情况
  • 任何非标准frontmatter字段

Phase 2: Individual Assessment

第二阶段：单个Skill评估

Evaluate each skill against these criteria:
根据以下标准评估每个Skill：

Frontmatter Quality

Frontmatter 质量

CriterionCheck
Description completenessDoes it explain both WHAT and WHEN?
Trigger coverageAre all reasonable trigger phrases included?
Description lengthSufficient detail without being excessive (50-300 words ideal)
Non-standard fieldsAny fields beyond name/description? (often unnecessary)
评估标准检查项
描述完整性是否同时说明了Skill的用途及适用场景？
触发词覆盖是否包含了所有合理的触发短语？
描述长度内容是否详实且不过度冗长（理想长度为50-300词）
非标准字段是否存在名称/描述之外的字段？（通常无必要）

Content Quality

内容质量

CriterionCheck
Line countUnder 500 lines? (Over suggests need for reference split)
Progressive disclosureDoes it use reference files for detailed content?
ActionabilityAre instructions concrete and procedural?
Output contractsDoes it specify expected output format?
ExamplesDoes it include good/bad examples where helpful?
评估标准检查项
行数是否少于500行？（超过则建议拆分至参考文件）
渐进式披露是否将详细内容存入参考文件？
可落地性操作说明是否具体、符合流程？
输出约定是否明确了预期的输出格式？
示例是否在必要时包含了正反示例？

Structure Quality

结构质量

CriterionCheck
Reference organizationAre references in 
references/
directory?
Script availabilityAre reusable operations scripted?
Internal linksAre reference files linked from SKILL.md?
Naming consistencyDo file names follow conventions?
评估标准检查项
参考文件组织参考文件是否存放在
references/
目录下？
脚本可用性可复用操作是否已脚本化？
内部链接SKILL.md中是否链接了参考文件？
命名一致性文件名称是否遵循统一规范？

Phase 3: Ecosystem Assessment

第三阶段：生态系统评估

Evaluate the skill collection as a whole:
从整体层面评估Skill集合：

Coverage Analysis

覆盖范围分析

  • Domain clusters: What areas do skills cover? Any gaps?
  • Overlap detection: Do multiple skills cover similar territory?
  • Workflow completeness: Can common tasks be handled end-to-end?
  • 领域集群：Skill覆盖了哪些领域？是否存在缺口？
  • 重叠检测：是否有多个Skill覆盖相似领域？
  • 工作流完整性：常见任务是否能端到端处理？

Consistency Analysis

一致性分析

  • Style uniformity: Do skills use consistent formatting?
  • Trigger pattern consistency: Are similar skills triggered similarly?
  • Output format consistency: Do related skills produce compatible outputs?
  • 风格统一性：Skill是否使用统一的格式？
  • 触发模式一致性：相似Skill的触发方式是否一致？
  • 输出格式一致性：相关Skill的输出是否兼容？

Organization Analysis

组织方式分析

  • Global vs project split: Are skills appropriately scoped?
  • Naming conventions: Are names consistent and discoverable?
  • Documentation quality: Is the overall system documented?
  • 全局与项目级拆分：Skill的范围划分是否合理？
  • 命名规范：名称是否统一且易于查找？
  • 文档质量：整个系统是否有完善的文档？

Phase 4: Generate Report

第四阶段：生成报告

Output Format

输出格式

markdown
undefined
markdown
undefined

Skill Ecosystem Audit

Skill Ecosystem Audit

Generated: [timestamp] Skills audited: [count] global, [count] project-level
Generated: [timestamp] Skills audited: [count] global, [count] project-level

Executive Summary

Executive Summary

[2-3 sentences: overall health, biggest opportunities, critical issues if any]
Health Score: [X/100]
[2-3 sentences: overall health, biggest opportunities, critical issues if any]
Health Score: [X/100]

Quick Stats

Quick Stats

MetricValueAssessment
Total skills[n]—
Avg SKILL.md lines[n][Good/High/Concerning]
Skills over 500 lines[n][Needs attention if >0]
Skills with references[n]%[Good if >30%]
Skills with scripts[n]%—
Duplicate coverage areas[n][Needs review if >0]
MetricValueAssessment
Total skills[n]—
Avg SKILL.md lines[n][Good/High/Concerning]
Skills over 500 lines[n][Needs attention if >0]
Skills with references[n]%[Good if >30%]
Skills with scripts[n]%—
Duplicate coverage areas[n][Needs review if >0]

Top Recommendations

Top Recommendations

Priority 1: [Most impactful improvement]

Priority 1: [Most impactful improvement]

[Specific action with affected skills]
[Specific action with affected skills]

Priority 2: [Second improvement]

Priority 2: [Second improvement]

[Specific action with affected skills]
[Specific action with affected skills]

Priority 3: [Third improvement]

Priority 3: [Third improvement]

[Specific action with affected skills]
[Specific action with affected skills]

Individual Skill Assessments

Individual Skill Assessments

[skill-name] — [Grade: A/B/C/D]

[skill-name] — [Grade: A/B/C/D]

Strengths:
  • [What works well]
Issues:
  • [Specific problem] → [Specific fix]
Quick wins:
  • [Actionable improvement]
[Repeat for each skill, ordered by grade (worst first for attention)]
Strengths:
  • [What works well]
Issues:
  • [Specific problem] → [Specific fix]
Quick wins:
  • [Actionable improvement]
[Repeat for each skill, ordered by grade (worst first for attention)]

Domain Coverage Map

Domain Coverage Map

DomainSkillsCoverage
[e.g., Design][list][Complete/Partial/Gap]
[e.g., Code Quality][list][Complete/Partial/Gap]
DomainSkillsCoverage
[e.g., Design][list][Complete/Partial/Gap]
[e.g., Code Quality][list][Complete/Partial/Gap]

Overlap Analysis

Overlap Analysis

[If overlaps exist, describe which skills overlap and recommend consolidation]
[If overlaps exist, describe which skills overlap and recommend consolidation]

Ecosystem Recommendations

Ecosystem Recommendations

Organization

Organization

  • [Recommendations for how skills are organized]
  • [Recommendations for how skills are organized]

Consistency

Consistency

  • [Recommendations for standardizing across skills]
  • [Recommendations for standardizing across skills]

Missing Skills

Missing Skills

  • [Suggestions for skills that would complement the ecosystem]
undefined
  • [Suggestions for skills that would complement the ecosystem]
undefined

Grading Rubric

评分标准

Individual Skills

单个Skill评分

GradeCriteria
AClear triggers, concise content (<300 lines), uses references appropriately, has output contract, well-organized
BGood triggers, reasonable length, mostly organized, minor improvements possible
CTriggers could be better, content somewhat long, organization issues, needs work
DWeak triggers, bloated content (>500 lines), poor organization, significant rewrite needed
等级标准
A触发词清晰、内容简洁（少于300行）、合理使用参考文件、有输出约定、组织有序
B触发词良好、长度合理、整体组织有序、仅需小幅优化
C触发词有待优化、内容偏长、存在组织问题、需要改进
D触发词薄弱、内容冗余（超过500行）、组织混乱、需要大幅重写

Ecosystem Health Score

生态系统健康评分

Calculate from:
  • Average individual grade (40%)
  • Coverage breadth without overlap (20%)
  • Consistency across skills (20%)
  • Progressive disclosure usage (10%)
  • Script/automation presence (10%)
评分计算依据：
  • 单个Skill平均得分（40%）
  • 无重叠的覆盖广度（20%）
  • Skill间的一致性（20%）
  • 渐进式披露的使用率（10%）
  • 脚本/自动化的覆盖率（10%）

Common Issues & Fixes

常见问题及修复方案

"Description doesn't include trigger phrases"

"描述未包含触发短语"

Fix: Add explicit "Use when..." phrases that match how users actually invoke the skill.
修复方案：添加明确的“适用场景：...”短语，匹配用户实际调用Skill的方式。

"SKILL.md over 500 lines"

"SKILL.md超过500行"

Fix: Move detailed content to 
references/
files. Keep only workflow and essential quick-reference in SKILL.md.
修复方案：将详细内容移至
references/
目录下的文件中。SKILL.md仅保留工作流及必要的快速参考内容。

"No output contract"

"无输出约定"

Fix: Add a section specifying what format the skill's output should take.
修复方案：添加一个章节，明确Skill输出应采用的格式。

"Duplicate coverage with [other-skill]"

"与[其他Skill]覆盖范围重复"

Fix: Either consolidate into one skill, or clearly differentiate triggers so they don't compete.
修复方案：要么合并为一个Skill，要么明确区分触发词，避免冲突。

"Missing reference links"

"缺少参考文件链接"

Fix: If references exist, link them from SKILL.md with guidance on when to load each.
修复方案：若存在参考文件，在SKILL.md中添加链接，并说明何时加载每个文件。

"Non-standard frontmatter fields"

"非标准frontmatter字段"

Fix: Remove fields beyond 
name
and 
description
unless specifically needed for skill behavior (rare).
修复方案：删除名称和描述之外的字段，除非Skill的运行逻辑确实需要（这种情况很少见）。

Proactive Recommendations

主动优化建议

Based on common skill ecosystem patterns, also check for:
  1. Missing meta-skill: Is there a skill for creating/improving skills? (skill-creator)
  2. Missing checkpoint skill: Is there a way to pause and assess? (checkpoint)
  3. Missing handoff skill: Can sessions be continued? (bootstrap)
  4. Missing review skills: Are there skills for quality gates? (code review, design critique)
基于常见的Skill生态系统模式，还需检查以下内容：
  1. 缺失元Skill：是否有用于创建/优化Skill的Skill？（如skill-creator）
  2. 缺失检查点Skill：是否有暂停并评估的方式？（如checkpoint）
  3. 缺失交接Skill：会话是否可以延续？（如bootstrap）
  4. 缺失审核类Skill：是否有用于质量把关的Skill？（如代码审核、设计评审）

Running the Audit

运行审核

Quick audit (overview only)

快速审核（仅概览）

bash
undefined
bash
undefined

Count skills and sizes

Count skills and sizes

find ~/.claude/skills -name "SKILL.md" -exec wc -l {} ; | sort -n find ./skills -name "SKILL.md" -exec wc -l {} ; 2>/dev/null | sort -n
undefined
find ~/.claude/skills -name "SKILL.md" -exec wc -l {} ; | sort -n find ./skills -name "SKILL.md" -exec wc -l {} ; 2>/dev/null | sort -n
undefined

Full audit

完整审核

  1. Inventory all skills
  2. Read each SKILL.md frontmatter and assess
  3. Check for reference/script directories
  4. Compile findings into report format above
  1. 整理所有Skill的清单
  2. 读取每个SKILL.md的frontmatter并进行评估
  3. 检查参考文件/脚本目录
  4. 将检查结果整理为上述报告格式

When to Audit

审核时机

  • After adding several new skills: Check for overlap and consistency
  • Periodically (quarterly): Ensure ecosystem stays healthy
  • Before sharing skills: Verify quality before distribution
  • When skills feel ineffective: Diagnose what's not working
  • 添加多个新Skill后：检查是否存在重叠及一致性问题
  • 定期（每季度）：确保生态系统保持健康
  • 分享Skill前：在分发前验证质量
  • Skill效果不佳时：诊断问题所在