agentsclimarketplace

Workflow metrics

Skill khanhhuyenngo985-sys/character-scene-design-skills/skills/.archive/ECC-codegen/workflow-metrics

创意工作流量化评分系统,评估项目进度和质量指标。触发词:「评分」「量化」「进度」「质量指标」。From its SKILL.md

Install
npx -y skills add khanhhuyenngo985-sys/character-scene-design-skills --skill workflow-metrics

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 5 stars5 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

3.8 KB, ~1.4k tokens by cl100k_base, as published. Nobody here has run it

创意工作流量化评分系统

功能

自动评估创意工作流的执行质量,输出可量化的指标分数和改进建议。

评估触发时机

  • 每个 Phase 完成后
  • 项目交付时
  • 每周自动汇总

核心指标

指标描述目标计算方式
decision_log_completeness决策日志完整率>95%实际记录数/应记录数
discussion_rounds_avg平均讨论轮次<3轮总轮次/讨论次数
reference_chain_coverage引用链覆盖率100%有引用链的工件数/总工件数
skill_routing_accuracy技能路由准确率>80%用户接受次数/推荐次数
quality_gate_first_pass质量门首次通过率>70%首次通过/总提交
user_intervention_rate用户干预率<30%干预次数/总阶段

评分计算方法

单项得分转换

decision_log_completeness: raw * 100
reference_chain_coverage: raw * 100
skill_routing_accuracy: raw * 100
quality_gate_first_pass: raw * 100
user_intervention_rate: (1 - raw) * 100  # 越低越好

discussion_rounds_avg: max(0, (5 - raw) / 5 * 100)  # 越少越好,<3轮得满分

权重

{
  "decision_log_completeness": 0.2,
  "discussion_rounds_avg": 0.15,
  "reference_chain_coverage": 0.2,
  "skill_routing_accuracy": 0.15,
  "quality_gate_first_pass": 0.2,
  "user_intervention_rate": 0.1
}

综合得分

overall = sum(single_score[k] * weight[k] for k in weight)
return round(overall, 1)  # 0-100

评分等级

等级分数说明
A90-100优秀,符合最佳实践
B80-89良好,有小幅改进空间
C70-79一般,存在明显问题
D60-69较差,需要重大改进
F<60不合格,需立即整改

输出格式

JSON 格式

{
  "project": "[项目名]",
  "evaluated_at": "[ISO时间戳]",
  "phase": "[当前阶段]",
  "metrics": {
    "decision_log_completeness": 0.95,
    "discussion_rounds_avg": 2.3,
    "reference_chain_coverage": 1.0,
    "skill_routing_accuracy": 0.85,
    "quality_gate_first_pass": 0.75,
    "user_intervention_rate": 0.2
  },
  "scores": {
    "decision_log_completeness": 95.0,
    "discussion_rounds_avg": 54.0,
    "reference_chain_coverage": 100.0,
    "skill_routing_accuracy": 85.0,
    "quality_gate_first_pass": 75.0,
    "user_intervention_rate": 80.0
  },
  "overall_score": 85.5,
  "grade": "B",
  "issues": [
    { "metric": "discussion_rounds_avg", "severity": "medium", "detail": "Phase 3 讨论了4轮" }
  ],
  "recommendations": [
    "建议在 Phase 3 增加预讨论,减少正式讨论轮次"
  ]
}

人类可读格式

## 项目评分报告:[项目名]

**综合得分:85.5 (B)**

### 指标详情
| 指标 | 数值 | 得分 |
|------|------|------|
| 决策日志完整率 | 95% | 95.0 |
| 平均讨论轮次 | 2.3轮 | 54.0 |
| 引用链覆盖率 | 100% | 100.0 |
| 技能路由准确率 | 85% | 85.0 |
| 质量门首次通过率 | 75% | 75.0 |
| 用户干预率 | 20% | 80.0 |

### 问题
- [中] Phase 3 讨论了4轮,建议增加预讨论

### 优化建议
1. 建议在 screenwriter agent 指令中增加决策日志检查点

使用方法

  1. 在 Phase 结束时,AI 自动运行评分
  2. 读取 project_state.json 和 decision_log.md
  3. 计算各项指标
  4. 输出评分报告
  5. 如有严重问题(等级D/F),暂停项目并要求整改

自动触发

在以下节点自动触发评分:

  • Phase 2 完成后(创意方向确认)
  • Phase 3-4 完成后(前期创作完成)
  • Phase 5 完成后(剧本定稿)
  • Phase 6 完成后(质量门通过)
  • Phase 8 完成后(项目交付)

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Keep looking

Skills are one crate of 326,679. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.