A professional skill auditing tool that scores Hermes Agent / Claude Code skills across 8 dimensions.
Supports GitHub URL or local directory, outputs terminal summary + rich HTML report.
# Audit any skill on GitHub
python review_tool.py https://github.com/lamwon/hermes-image-generation-skill --html
# Open the report
open skill-review-report.htmlYou get:
代码质量 [####------] 4.0/10
文档完整性 [#########-] 9.0/10
...
>>> 总分: 7.0/10 评级: B <<<
改进建议:
1. 增加 try/except 错误处理
2. 添加 --help 参数说明
...
Plus a full HTML report with:
- Repo info (stars, topics, description)
- File structure overview
- 8-dimension bar chart with color-coded scores
- Total score + letter grade (S/A/B/C/D)
- Sorted improvement suggestions
git clone https://github.com/lamwon/hermes-skill-review.git
cd hermes-skill-review# Terminal summary only
python review_tool.py https://github.com/lamwon/hermes-image-generation-skill
# With HTML report
python review_tool.py https://github.com/lamwon/hermes-image-generation-skill --html
# Custom output path
python review_tool.py https://github.com/user/repo --html -o my-report.html
# Local directory
python review_tool.py /path/to/skill/folder --htmlhermes skills install lamwon/hermes-skill-reviewThen in Hermes:
/review-skill https://github.com/lamwon/hermes-image-generation-skill
Or in conversation:
帮我审查这个技能:https://github.com/xxx/skill-name
| # | Dimension | Weight | What it checks |
|---|---|---|---|
| 1 | Code Quality | 1.0x | Syntax correctness, error handling, security, style |
| 2 | Documentation | 1.0x | README/SKILL.md completeness, install guide, examples, FAQ |
| 3 | Architecture | 0.8x | File structure, dependency management, engineering config |
| 4 | User Experience | 0.9x | Ease of install, intuitive args, Chinese-friendly output |
| 5 | Error Handling | 0.8x | Retry logic, rate limit handling, edge cases |
| 6 | Portability | 0.6x | Cross-platform (Windows/Mac/Linux), encoding, paths |
| 7 | Maintainability | 0.6x | Comments, centralized config, versioning |
| 8 | Innovation | 0.5x | Unique value, differentiation, comparison tables |
Grading: S (9-10) > A (8-9) > B (7-8) > C (5-7) > D (<5)
The HTML report features a dark theme with:
- Color-coded rating badge (S/A/B/C/D)
- Repo metadata card
- File structure tags
- Animated score bars
- Actionable improvement suggestions
- Print-friendly layout
- Python 3.6+ (standard library only — no pip install needed)
- Internet access (for GitHub URL mode only)
| Who | When to use |
|---|---|
| Skill authors | Before publishing to get objective feedback |
| Skill reviewers | To quickly evaluate PRs / submissions |
| Team leads | To enforce quality standards across skills |
| Learners | To understand what makes a good skill |
MIT