Skill Reviewer
Comprehensive skill review and validation for OpenClaw skills.
Overview
This skill provides a systematic three-level review process to ensure skills are:
- - Valid - Proper structure and format
- Complete - All required components present
- Correct - Generated outputs match OpenClaw specifications
- High-quality - Follows best practices
Review Workflow
Level 1: Format Validation
Run automatic validation using package_skill.py:
CODEBLOCK0
Checks:
- - YAML frontmatter format and required fields
- Skill naming conventions and directory structure
- Description completeness and quality
- File organization and resource references
Outcome: ✅ Valid or ❌ Validation errors
Level 2: Content Quality Assessment
Manually review SKILL.md for:
Frontmatter:
- -
name is concise and follows naming conventions - INLINECODE2 is comprehensive and includes "when to use" information
- No extraneous fields in YAML
Body Structure:
- - Clear organization (workflow-based, task-based, or reference-based)
- Progressive disclosure pattern (metadata → SKILL.md → references)
- Concise and focused content
- Imperative/infinitive form for instructions
Writing Quality:
- - No filler or redundant explanations
- Concrete examples and realistic scenarios
- Clear guidance on when to read reference files
- Avoids duplication between SKILL.md and references
Level 3: Functional Verification
Critical Step - Verify that generated outputs match actual OpenClaw specifications.
For skills that generate templates:
- - Compare generated templates with actual OpenClaw specification files
- Example: If skill generates AGENTS.md templates, compare with INLINECODE3
- Check for missing required sections (session startup, memory workflow, safety rules, group chat etiquette, heartbeat mechanism)
- Verify all critical requirements are present
For skills with scripts:
- - Test scripts to ensure they work correctly
- Verify output matches expected format
- Check error handling
For skills with references:
- - Verify reference files are accurate and up-to-date
- Check that references are properly linked from SKILL.md
Level 4: Best Practices Check
Verify the skill follows OpenClaw skill best practices:
Progressive Disclosure:
- - SKILL.md body is concise (<500 lines preferred)
- Detailed information moved to references/
- References are properly linked and described
Resource Organization:
- - Only necessary resource directories created
- No extraneous files (README.md, INSTALLATION_GUIDE.md, etc.)
- Scripts/ references/ assets/ used appropriately
Context Efficiency:
- - Information lives in either SKILL.md OR references, not both
- Essential procedural instructions in SKILL.md
- Detailed reference material in references/
Triggering Accuracy:
- -
description clearly states when the skill should be used - All "when to use" information is in description, not body
Common Issues Found
Missing Functional Verification:
- - Skill generates templates but they don't match actual specifications
- Example: AGENTS.md template missing session startup requirements
Incomplete Descriptions:
- - Description doesn't include "when to use" information
- Body contains "When to Use This Skill" sections (should be in description)
Duplication:
- - Same information in both SKILL.md and references
- Wastes context window tokens
Extraneous Files:
- - README.md, CHANGELOG.md, etc. in skill directory
- Should only contain SKILL.md and necessary resources
Poor Progressive Disclosure:
- - SKILL.md too verbose
- References not properly linked or described
Review Checklist
For each skill, verify:
- - [ ] Level 1: Format validation passes
- [ ] Level 2: SKILL.md structure and quality are good
- [ ] Level 3: Generated templates match OpenClaw specifications
- [ ] Level 4: Best practices are followed
- [ ] No extraneous files
- [ ] Description includes "when to use" information
- [ ] References are properly linked from SKILL.md
- [ ] Scripts work correctly (if present)
Resources
scripts/
- -
review_skill.py - Automated review script (optional enhancement)
references/
- -
openclaw-specs.md - OpenClaw specifications for comparison - INLINECODE7 - Skill design best practices
技能审查器
针对OpenClaw技能的系统性审查与验证。
概述
本技能提供系统化的三级审查流程,确保技能具备以下特性:
- - 有效性 - 结构格式正确
- 完整性 - 包含所有必要组件
- 正确性 - 生成输出符合OpenClaw规范
- 高质量 - 遵循最佳实践
审查工作流
第一级:格式验证
使用package_skill.py运行自动验证:
bash
python3 /home/yupeng/.npm-global/lib/nodemodules/openclaw/skills/skill-creator/scripts/packageskill.py <技能路径>
检查项:
- - YAML前置元数据格式及必填字段
- 技能命名规范与目录结构
- 描述完整性与质量
- 文件组织与资源引用
结果: ✅ 通过 或 ❌ 验证错误
第二级:内容质量评估
手动审查SKILL.md文件:
前置元数据:
- - name字段简洁且符合命名规范
- description字段全面且包含使用时机信息
- YAML中无多余字段
主体结构:
- - 清晰的组织方式(基于工作流、任务或参考)
- 渐进式信息呈现模式(元数据 → SKILL.md → 参考文件)
- 内容简洁聚焦
- 指令采用祈使句/不定式形式
写作质量:
- - 无冗余解释或填充内容
- 包含具体示例和真实场景
- 明确指导何时阅读参考文件
- 避免SKILL.md与参考文件内容重复
第三级:功能验证
关键步骤 - 验证生成输出与实际OpenClaw规范是否匹配。
对于生成模板的技能:
- - 将生成的模板与实际OpenClaw规范文件对比
- 示例:若技能生成AGENTS.md模板,需与/home/yupeng/.openclaw/workspace/AGENTS.md对比
- 检查是否缺少必要章节(会话启动、记忆工作流、安全规则、群聊礼仪、心跳机制)
- 验证所有关键需求均已包含
对于包含脚本的技能:
- - 测试脚本确保正常运行
- 验证输出格式符合预期
- 检查错误处理机制
对于包含参考文件的技能:
- - 验证参考文件准确且为最新版本
- 检查参考文件是否在SKILL.md中正确链接
第四级:最佳实践检查
验证技能是否遵循OpenClaw技能最佳实践:
渐进式信息呈现:
- - SKILL.md主体内容简洁(建议少于500行)
- 详细信息移至references/目录
- 参考文件正确链接并说明
资源组织:
- - 仅创建必要的资源目录
- 无多余文件(README.md、INSTALLATION_GUIDE.md等)
- 合理使用scripts/、references/、assets/目录
上下文效率:
- - 信息仅存在于SKILL.md或参考文件中,避免重复
- 关键操作说明放在SKILL.md
- 详细参考资料放在references/
触发准确性:
- - description字段明确说明技能使用时机
- 所有使用时机信息均在描述中,而非主体内容
常见问题
功能验证缺失:
- - 技能生成模板但与实际规范不匹配
- 示例:AGENTS.md模板缺少会话启动需求
描述不完整:
- - 描述未包含使用时机信息
- 主体内容包含何时使用本技能章节(应放在描述中)
内容重复:
- - SKILL.md与参考文件包含相同信息
- 浪费上下文窗口令牌
多余文件:
- - 技能目录中存在README.md、CHANGELOG.md等文件
- 应仅包含SKILL.md和必要资源
渐进式信息呈现不佳:
- - SKILL.md内容过于冗长
- 参考文件未正确链接或说明
审查清单
针对每个技能,验证以下项目:
- - [ ] 第一级:格式验证通过
- [ ] 第二级:SKILL.md结构与质量良好
- [ ] 第三级:生成模板与OpenClaw规范匹配
- [ ] 第四级:遵循最佳实践
- [ ] 无多余文件
- [ ] 描述包含使用时机信息
- [ ] 参考文件在SKILL.md中正确链接
- [ ] 脚本运行正常(如存在)
资源
scripts/
- - review_skill.py - 自动化审查脚本(可选增强功能)
references/
- - openclaw-specs.md - 用于对比的OpenClaw规范
- best-practices.md - 技能设计最佳实践