$npx -y skills add xiaomoBoy/claude-writing-skills --skill score-optimizerUse when the user wants to iterate on a viral-article scoring system itself, calibrate or improve a scoring prompt against labeled samples, or run batch scoring experiments on a fixed article set. Best for prompt-only scoring research where the evaluator scripts stay fixed and on
| 1 | # Score Optimizer |
| 2 | |
| 3 | 这个 skill 只负责一件事:优化文章评分系统本身。 |
| 4 | |
| 5 | 它适合做的是: |
| 6 | |
| 7 | - 调整文章爆款评分 Prompt |
| 8 | - 批量给样本文集打分 |
| 9 | - 对照标注数据评估评分准确度 |
| 10 | - 研究评分维度、权重、few-shot 示例是否更合理 |
| 11 | |
| 12 | 它不负责: |
| 13 | |
| 14 | - 直接改文章正文 |
| 15 | - 平台改写 |
| 16 | - 发布包装 |
| 17 | |
| 18 | ## Working Scope |
| 19 | |
| 20 | 适用场景: |
| 21 | |
| 22 | - 用户说“优化文章评分 prompt” |
| 23 | - 用户说“让这个爆款评分系统更准” |
| 24 | - 用户说“跑一轮样本文集评分实验” |
| 25 | - 用户说“校准 viral score / composite score 的打分规则” |
| 26 | |
| 27 | 不适用场景: |
| 28 | |
| 29 | - 用户说“直接把这篇文章改高分” |
| 30 | - 用户说“按这个评分器去优化文章内容” |
| 31 | |
| 32 | 后者应改用 `article-optimizer`。 |
| 33 | |
| 34 | ## Required Reads |
| 35 | |
| 36 | 开始前先读: |
| 37 | |
| 38 | 1. [references/score_program.md](references/score_program.md) |
| 39 | 2. [references/score_prompt.md](references/score_prompt.md) |
| 40 | |
| 41 | 需要看实现时再读: |
| 42 | |
| 43 | - [scripts/run_scoring.py](scripts/run_scoring.py) |
| 44 | - [scripts/evaluate.py](scripts/evaluate.py) |
| 45 | - [assets/articles/labels.example.json](assets/articles/labels.example.json) —— 标注 schema 示例 |
| 46 | |
| 47 | 样本文集位置(**用户自己提供**,本 skill 不带样本): |
| 48 | |
| 49 | - `assets/articles/samples/` —— 把你自己标注过的 .md 文章放进来,然后 `cp labels.example.json labels.json` 填上你的评分。详见 [assets/articles/samples/README.md](assets/articles/samples/README.md) |
| 50 | |
| 51 | ## Workflow |
| 52 | |
| 53 | 1. 先确认当前任务是在优化评分器,不是在改文章。 |
| 54 | 2. 读取 `references/score_program.md`,按它的实验约束执行。 |
| 55 | 3. 默认只改 `references/score_prompt.md`,不要动评估脚本。 |
| 56 | 4. 运行批量评分脚本,再运行评估脚本。 |
| 57 | 5. 比较 `composite_score`、`mae`、`spearman`、`classification`。 |
| 58 | 6. 记录结果,决定保留还是回滚。 |
| 59 | |
| 60 | ## Commands |
| 61 | |
| 62 | 在 skill 根目录执行: |
| 63 | |
| 64 | ```bash |
| 65 | python3 scripts/run_scoring.py > scoring.log 2>&1 |
| 66 | python3 scripts/evaluate.py > eval.log 2>&1 |
| 67 | grep "^composite_score:\|^mae:\|^spearman:\|^classification:" eval.log |
| 68 | ``` |
| 69 | |
| 70 | 如果需要图表: |
| 71 | |
| 72 | ```bash |
| 73 | python3 scripts/generate_score_chart.py |
| 74 | ``` |
| 75 | |
| 76 | ## Guardrails |
| 77 | |
| 78 | - 默认不要改 `scripts/evaluate.py` |
| 79 | - 默认不要改 `scripts/run_scoring.py` |
| 80 | - 默认不要安装新依赖 |
| 81 | - 如果只是想提高某一篇文章分数,不要用这个 skill |