feat: close the scoring feedback loop in pipeline

Step 8a: write composite_score + writing_config_snapshot to history.yaml,
recording which parameters produced which anti-AI score.

Step 4a-0: before writing, read history for the best-scoring article's
parameter combination and use it as reference for the current article.

This closes the feedback loop: write → score → record → learn → write better.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
wangzhuc 2026-03-30 20:19:57 +08:00
parent 70bfca61b0
commit a98a99e773

View file

@ -177,10 +177,17 @@ WebSearch: "{选题关键词} 数据 报告 2025 2026"
```
读取: {skill_dir}/references/writing-guide.md
读取: {skill_dir}/playbook.md如果存在逐条执行优先于 writing-guide
读取: {skill_dir}/writing-config.yaml如果存在作为写作参数
读取: {skill_dir}/history.yaml最近 3 篇的 dimensions 字段)
```
**4a. 维度随机化**:从 writing-guide.md 第 7 层维度池随机激活 2-3 个维度,对比历史去重。
**4a-0. 历史最佳参数参考**(有 history.yaml 且包含 composite_score 时执行):
读取 history.yaml 中有 `composite_score``writing_config_snapshot` 的文章,找到得分最低(最人类)的一篇。如果该篇得分比当前 writing-config.yaml 的默认参数对应的历史平均分更好,在写作时**参考**其参数组合(不是覆盖 writing-config.yaml而是作为"上次这组参数效果好"的提示)。
具体:如果历史最佳文章的某个参数值与当前 writing-config 不同,在写作时倾向使用历史最佳值。如果没有历史数据,跳过此步。
**4a. 维度随机化**:从 writing-guide.md 维度池随机激活 2-3 个维度,对比历史去重。
**4b. 加载写作人格**
@ -315,6 +322,16 @@ python3 {skill_dir}/toolkit/cli.py preview {markdown} --theme {theme} --no-open
writing_persona: "{人格名}"
dimensions:
- "{维度}: {选项}"
composite_score: {Step 5b-2 的 composite_score} # 0=人类, 100=AI
writing_config_snapshot: # 本次使用的关键参数(从 writing-config.yaml 提取)
sentence_variance: {值}
paragraph_rhythm: "{值}"
emotional_arc: "{值}"
word_temperature_bias: "{值}"
broken_sentence_rate: {值}
tangent_frequency: "{值}"
style_drift: {值}
negative_emotion_floor: {值}
stats: null
```