← 返回 Skills 市场
105
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install skill-review-ai
功能描述
审查 Agent Skills 的规范性、完整性和代码质量。在安装或发布 skills 时使用,验证 SKILL.md 格式、目录结构、脚本代码和文件引用是否符合 Agent Skills 规范。
使用说明 (SKILL.md)
Skill Review
审查 Agent Skills 的规范性、完整性和代码质量。
使用场景
- 安装 skills 前进行质量检查
- 发布 skills 前的合规验证
- 代码审查和最佳实践检查
- 目录结构和文件引用验证
审查项目
1. SKILL.md 格式审查
- YAML frontmatter 存在且格式正确
-
name字段符合命名规范(小写字母、数字、连字符,1-64字符) -
description字段存在且长度在 1-1024 字符之间 -
license字段(如有)格式正确 -
compatibility字段(如有)长度在 500 字符以内 -
metadata字段(如有)为键值对格式 -
allowed-tools字段(如有)格式正确
2. 目录结构审查
- 技能目录名与
name字段一致 -
SKILL.md文件存在于根目录 -
scripts/目录(如有)结构合理 -
references/目录(如有)结构合理 -
assets/目录(如有)结构合理 - 无深层嵌套(建议不超过 2 层)
3. 脚本代码审查
- 脚本文件有执行权限(如需要)
- 脚本包含清晰的错误处理
- 脚本有适当的注释说明
- 脚本依赖关系明确
- 无硬编码的敏感信息
- 代码风格一致
4. 文件引用审查
- 所有引用的文件存在
- 相对路径正确(从 skill 根目录开始)
- 无循环引用
- 引用的文件大小合理
使用方法
基本用法
# 审查单个 skill
bash scripts/review.sh /path/to/skill-name
# 详细审查(包含代码分析)
bash scripts/review.sh /path/to/skill-name --verbose
# 生成 JSON 格式报告
bash scripts/review.sh /path/to/skill-name --json
Python API
from scripts.review_skill import SkillReviewer
reviewer = SkillReviewer()
result = reviewer.review("/path/to/skill-name")
print(result.to_markdown())
审查报告格式
Markdown 报告
# Skill 审查报告: skill-name
## 概要
- 总分: 85/100
- 状态: ✅ 通过 / ❌ 未通过
- 审查时间: 2024-01-01 12:00:00
## 详细结果
### SKILL.md 格式
- 状态: ✅ 通过
- 得分: 25/25
- 问题: 无
### 目录结构
- 状态: ✅ 通过
- 得分: 20/20
- 问题: 无
### 脚本代码
- 状态: ⚠️ 警告
- 得分: 30/35
- 问题:
- [WARN] 脚本缺少错误处理: scripts/process.py:15
- [WARN] 建议添加注释: scripts/helper.sh:8
### 文件引用
- 状态: ❌ 失败
- 得分: 10/20
- 问题:
- [ERROR] 引用的文件不存在: references/MISSING.md
- [WARN] 文件过大: assets/large-file.bin (5.2MB)
JSON 报告
{
"skill_name": "skill-name",
"overall_score": 85,
"status": "passed",
"timestamp": "2024-01-01T12:00:00Z",
"categories": {
"skill_md": {
"score": 25,
"max_score": 25,
"status": "passed",
"issues": []
},
"directory_structure": {
"score": 20,
"max_score": 20,
"status": "passed",
"issues": []
},
"script_code": {
"score": 30,
"max_score": 35,
"status": "warning",
"issues": [
{
"level": "warn",
"message": "脚本缺少错误处理",
"file": "scripts/process.py",
"line": 15
}
]
},
"file_references": {
"score": 10,
"max_score": 20,
"status": "failed",
"issues": [
{
"level": "error",
"message": "引用的文件不存在",
"file": "references/MISSING.md"
}
]
}
}
}
评分标准
| 类别 | 满分 | 通过线 | 说明 |
|---|---|---|---|
| SKILL.md 格式 | 25 | 20 | frontmatter 和正文格式 |
| 目录结构 | 20 | 15 | 目录组织和命名 |
| 脚本代码 | 35 | 25 | 代码质量和安全性 |
| 文件引用 | 20 | 15 | 引用完整性和正确性 |
| 总分 | 100 | 75 | 综合评分 |
常见问题
SKILL.md 问题
问题: name 字段包含大写字母
# ❌ 错误
name: PDF-Processing
# ✅ 正确
name: pdf-processing
问题: description 过短
# ❌ 错误
description: Helps with PDFs.
# ✅ 正确
description: Extracts text and tables from PDF files, fills PDF forms, and merges multiple PDFs. Use when working with PDF documents.
目录结构问题
问题: 目录名与 name 字段不匹配
# ❌ 错误
my-skill/
SKILL.md (name: different-name)
# ✅ 正确
my-skill/
SKILL.md (name: my-skill)
脚本问题
问题: 缺少错误处理
# ❌ 错误
rm -rf $TARGET_DIR
# ✅ 正确
if [ -d "$TARGET_DIR" ]; then
rm -rf "$TARGET_DIR" || { echo "无法删除目录"; exit 1; }
fi
文件引用问题
问题: 引用的文件不存在
# ❌ 错误
See [reference](references/NONEXISTENT.md)
# ✅ 正确
See [reference](references/EXISTING.md)
参考文档
最佳实践
- 在安装 skill 前运行审查: 确保 skill 质量符合预期
- 在 CI/CD 中集成: 自动化 skill 发布流程
- 定期更新审查规则: 跟随 Agent Skills 规范更新
- 结合人工审查: 自动化审查不能替代人工判断
安全使用建议
This appears to be a coherent, read-only skill-audit tool. Before running it: 1) Only point it at skill directories you control or trust, since it will read all files in the supplied path. 2) The tool requires Python to run and optionally benefits from yamllint/jq; missing optional tools only reduce checks. 3) The reviewer does not request credentials or network access, but as with any script, review the code if you have high security requirements. If you need higher assurance, inspect scripts/review_skill.py and review.sh locally (they are included) to confirm no unexpected behavior.
功能分析
Type: OpenClaw Skill
Name: skill-review-ai
Version: 1.0.0
The skill bundle is a legitimate utility designed to perform static analysis and quality checks on other Agent Skills. The core logic in `scripts/review_skill.py` and `scripts/review.sh` focuses on verifying directory structures, metadata formatting, and identifying potential security risks like hardcoded secrets (passwords, API keys) using regular expressions, with no evidence of malicious intent, data exfiltration, or unauthorized execution.
能力评估
Purpose & Capability
Name, description, included scripts (review.sh, review_skill.py), and documentation align with a skill-auditing tool; required binaries/env are none and any optional dependencies (yamllint, jq, python3) match expected functionality.
Instruction Scope
SKILL.md instructs the agent to run local review scripts and the Python API to analyze a skill directory. The reviewer reads files under the target skill path (SKILL.md, scripts/, references/, etc.), which is expected for this purpose — it does not contain instructions to exfiltrate data or call external endpoints. Users should note the tool inspects all files in the provided directory (read-only behavior is claimed in docs).
Install Mechanism
No install spec (instruction-only with bundled scripts). No downloads from external URLs or archive extraction; included code runs locally. This is the lowest-risk install pattern.
Credentials
No environment variables, credentials, or config paths are required. The scripts warn about optional tooling (yamllint, jq) but do not require unrelated secrets.
Persistence & Privilege
always:false and agent-autonomy left at default. The skill does not request permanent presence or modify other skills or global agent config; it runs as a normal, user-invoked tool.
如何使用
- 确保已安装 OpenClaw(本地或 Docker 部署)
- 在对话框中输入安装命令:
/install skill-review-ai - 安装完成后,直接呼叫该 Skill 的名称或使用
/skill-review-ai触发 - 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.0
Initial release of skill-review-ai
- Provides a comprehensive checklist for reviewing Agent Skills for specification, completeness, and code quality.
- Details requirements and best practices for SKILL.md format, directory structure, scripts, and file references.
- Outlines report formats (Markdown and JSON) and scoring standards.
- Provides usage instructions for both CLI and Python API.
- Includes example issues and common problems to aid in troubleshooting.
- Lists references and best practices for continuous integration and manual review.
元数据
常见问题
skill审查 是什么?
审查 Agent Skills 的规范性、完整性和代码质量。在安装或发布 skills 时使用,验证 SKILL.md 格式、目录结构、脚本代码和文件引用是否符合 Agent Skills 规范。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 105 次。
如何安装 skill审查?
在 OpenClaw 或 Claude Code 对话框中运行命令「/install skill-review-ai」即可一键安装,无需额外配置。
skill审查 是免费的吗?
是的,skill审查 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。
skill审查 支持哪些平台?
skill审查 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。
谁开发了 skill审查?
由 rumu14(@hwl1413520)开发并维护,当前版本 v1.0.0。
推荐 Skills