skill-scorer

Files (92)

• SKILL.md
• USAGE.md
• rubric.yaml · domains/finance/
• SKILL.md · examples/banking-workflow-assistant/
• SKILL.md · examples/finance-scenario-advisor/
• SKILL.md · examples/financial-data-analysis-agent/
• SKILL.md · examples/financial-education-coach/
• SKILL.md · examples/mega-pipeline/
• SKILL.md · examples/mega-pipeline/agents/a11y/
• SKILL.md · examples/mega-pipeline/agents/api/
• SKILL.md · examples/mega-pipeline/agents/async/
• SKILL.md · examples/mega-pipeline/agents/audit/
• SKILL.md · examples/mega-pipeline/agents/authn/
• SKILL.md · examples/mega-pipeline/agents/authz/
• SKILL.md · examples/mega-pipeline/agents/backend/
• SKILL.md · examples/mega-pipeline/agents/caching/
• SKILL.md · examples/mega-pipeline/agents/circuit/
• SKILL.md · examples/mega-pipeline/agents/complexity/
• SKILL.md · examples/mega-pipeline/agents/compliance/
• SKILL.md · examples/mega-pipeline/agents/config/
• SKILL.md · examples/mega-pipeline/agents/cors/
• SKILL.md · examples/mega-pipeline/agents/coverage/
• SKILL.md · examples/mega-pipeline/agents/crypto/
• SKILL.md · examples/mega-pipeline/agents/csrf/
• SKILL.md · examples/mega-pipeline/agents/deps/
• SKILL.md · examples/mega-pipeline/agents/docs/
• SKILL.md · examples/mega-pipeline/agents/env/
• SKILL.md · examples/mega-pipeline/agents/events/
• SKILL.md · examples/mega-pipeline/agents/frontend/
• SKILL.md · examples/mega-pipeline/agents/gdpr/
• SKILL.md · examples/mega-pipeline/agents/headers/
• SKILL.md · examples/mega-pipeline/agents/i18n/
• SKILL.md · examples/mega-pipeline/agents/idor/
• SKILL.md · examples/mega-pipeline/agents/imports/
• SKILL.md · examples/mega-pipeline/agents/infra/
• SKILL.md · examples/mega-pipeline/agents/licensing/
• SKILL.md · examples/mega-pipeline/agents/logging/
• SKILL.md · examples/mega-pipeline/agents/memory/
• SKILL.md · examples/mega-pipeline/agents/metrics/
• SKILL.md · examples/mega-pipeline/agents/migration/
• SKILL.md · examples/mega-pipeline/agents/naming/
• SKILL.md · examples/mega-pipeline/agents/observability/
• SKILL.md · examples/mega-pipeline/agents/perf/
• SKILL.md · examples/mega-pipeline/agents/privacy/
• SKILL.md · examples/mega-pipeline/agents/ratelimit/
• SKILL.md · examples/mega-pipeline/agents/rbac/
• SKILL.md · examples/mega-pipeline/agents/retries/
• SKILL.md · examples/mega-pipeline/agents/schema/
• SKILL.md · examples/mega-pipeline/agents/secrets/
• SKILL.md · examples/mega-pipeline/agents/security/
• SKILL.md · examples/mega-pipeline/agents/serialization/
• SKILL.md · examples/mega-pipeline/agents/sql/
• SKILL.md · examples/mega-pipeline/agents/sqli/
• SKILL.md · examples/mega-pipeline/agents/streaming/
• SKILL.md · examples/mega-pipeline/agents/style/
• SKILL.md · examples/mega-pipeline/agents/tests/
• SKILL.md · examples/mega-pipeline/agents/threading/
• SKILL.md · examples/mega-pipeline/agents/timeout/
• SKILL.md · examples/mega-pipeline/agents/tracing/
• SKILL.md · examples/mega-pipeline/agents/typing/
• …and 32 more

SKILL.md (first lines)

# skill-scorer

一个"评测 Skill 的 Skill"。接收任意 Agent Skill 的源文件，必须依据本仓库的官方评分入口和 `rubric/rubric.yaml`
给出 5 大支柱的 100 分制评分、等级、证据引用与改进建议。Rubric 内置三类型差异化（atomic / pipeline / composite），子维度数随 skill 结构自动启用，由 `applies_to` 字段控制。同时兼容 **Cursor / Claude / OpenClaw** 三套规范。

## When to use

- 用户提供 `SKILL.md` / skill 文件夹 / `.zip` / GitHub URL，并请求评分、审计或改进建议。
- 用户询问"我这个 skill 写得怎么样"、"怎么提升我的 skill 质量"、"帮我对齐官方最佳实践"。
- **不适用于**：评价非 Skill 类文档（普通 README / 博客 / prompt 模板）。

## Code Agent Quick Start

如果你是 Cursor、WorkBuddy、Hermes、小龙虾或类似 code agent，先读 `USAGE.md`。

推荐先运行 CLI 向导，让用户选择通用评测或金融专家版；如果选择金融专家版，向导会继续确认金融子场景，并输出后续官方命令：

```bash
python3 skills/skill-scorer/scripts/score.py --agent-wizard <path-to-skill-zip-dir-or-SKILL.md>
```

规则分预览：

```bash
python3 skills/skill-scorer/scripts/score.py <path-to-skill-zip-dir-or-SKILL.md>
```

完整 agent-side Deep Review（使用 code agent 自己的模型套餐，不消耗 SkillLens 服务端 key）：

```bash
python3 skills/skill-scorer/scripts/score.py --agent-prompt <path-to-skill-zip-dir-or-SKILL.md> > agent-deep-review-prompt.md
# 将 agent-deep-review-prompt.md 完整交给当前 code agent 的模型，保存严格 JSON 为 agent-llm-results.json
python3 skills/skill-scorer/scripts/score.py --llm-results agent-llm-results.json <path-to-skill-zip-dir-or-SKILL.md>
```

不得临时生成自定义评分脚本替代官方 CLI；最终分数必须来自最后一步官方 CLI 输出。

金融专家版（可选）应优先通过 `--agent-wizard` 选择；手动执行时，必须在 `--agent-prompt` 和 `--llm-results` 两步都加入相同的 `--domain finance --scenario <scenario-id>`。支持的场景详见 `USAGE.md`。

## Inputs

- 一个 `SKILL.md` 文本，或
- 一个 skill 目录（含 `scripts/` `references/` `assets/` 等），或
- 一个 `.zip` 打包的 skill，或
- 一个指向 skill 仓库/子目录的 GitHub URL（Web 工具侧支持）。

## Outputs

```json
{
  "spec": "claude | openclaw",
  "language": "zh | en",
  "score": 0-100,
  "grade": "S | A | B | C | D",
  "pillars": [
    {
      "id": "business_value",
      "score": 0-25,
      "dimensions": [
        {
          "id": "...",
          "checks": [
            {
              "id": "...",
              "status": "pass|partial|fail|n_a",
              "evidence":    "<primary-language alias>",
              "evidence_zh": "中文现状",
              "evidence_en": "English diagnosis",
              "fix":    "<primary-language alias>",
              "fix_zh": "中文改法",
              "fix_en": "English fix"
            }
          ]
        }
      ]
    }
  ],
  "bonus": 0-5,
  "suggestions": [
    {
      "title": "Top 改进项",
      "title_zh": "中文 Top 改进项",
      "title_en": "English Top Improvement",
      "why":    "现状",
      "why_zh": "中文现状",
      "why_en": "English why",
      "how":    "改法",
      "how_zh": "中文改法",
      "how_en": "English how"
    }
  ],
  "deepReviewCertificate": {
    "status": "verified"
  }
}
```

`evidence_zh` + `evidence_en` (and `fix_zh` + `fix_en`, `why_zh` + `why_en`, `how_zh` + `how_en`, `title_zh` + `title_en`) are the canonical bilingual fields ≥ engineVersion 0.4.1. The unsuffixed `evidence` / `fix` / `why` / `how` / `title` are preserved as back-compat aliases pointing at the primary language so older readers keep working. The HTML report's ZH/EN toggle uses the suffixed fields to switch body content; falls back to the bare field when the JSON predates the bilingual schema.

## Workflow

1. **Locate SkillLens root**：先定位包含 `skills/skill-scorer/rubric/rubric.yaml` 的 SkillLens 仓库根目录。
2. **Run official scorer**：运行官方 CLI，不得临时生成替代评分脚本：

   ```bash
   python3 skills/skill-scorer/scripts/score.py <path-to-skill-zip-dir-or-SKILL.md>
   ```

3. **Choose review mode**：优先运行 `--agent-wizard`。如手动执行，必须确认是否启用领域专家版；当前 MVP 支持 `finance`，并必须确认具体 `--scenario`。
4. **Agent-side Deep Review when requested**：如需完整深度评测，必须先运行 `--agent-prompt` 生成官方提示词，用当前 code agent 的模型返回严格 JSON，再运行 `--llm-results` 合并。领域专家版必须在两步命令都带上相同的 `--domain` / `--scenario`。
5. **Use official JSON only**：总分、等级、pillar/dimension/check 分数必须来自官方 CLI 最终 JSON 输出，不能由 Agent 自己重算或补满。
6. **Verify certificate**：完整 Deep Review 必须包含 `deepReviewCertificate.status="verified"`；金融专家版还必须包含 `domainExpert` 和 `deepReviewCertificate.domain`；没有证书只能称为规则分预览或非官方结果。
7. **Render**：按用户阅读语言（zh / en）从 JSON 取双语字段（`evidence_zh` + `evidence_en`, `fix_zh` + `fix_en`, `why_zh` + `why_en`, `how_zh` + `how_en`）渲染报告；Top 改进项必须来自 JSON 的 `suggestions`，旧版单语 JSON 可回退到 `evidence` / `fix` / `why` / `how`。

## Official Tool Contract

- **MUST** call `skills/skill-scorer/scripts/score.py` for local tool use, or call the deployed SkillLens Web/API endpoint when the user explicitly提供该服务地址。
- **SHOULD** start with `--agent-wizard` for agent-side Deep Review so the user explicitly chooses general vs. finance expert review.
- **MUST** use the official `--agent-prompt` → model JSON → `--llm-results` flow for agent-side Deep Review.

From AndrewNgGirl/SkillLens, MIT.