| name | fix-text |
| version | 2.0.0 |
| description | 文本修正(AI腔+禁用词+标点) |
| category | fix |
| dual_mode | {"standalone":true,"orchestrator":true} |
| inputs | [{"name":"chapter_file","type":"file","required":true,"description":"章节文件路径","default_source":"正文/第{N}章.md"},{"name":"fix_type","type":"text","required":false,"description":"修正类型:ai|banned|punctuation|all(默认all)","default_value":"all"}] |
| outputs | [{"name":"fixed_text","type":"text"}] |
| script | _shared/scripts/punctuation-normalize.js |
| references | ["_shared/references/banned-words.md","_shared/references/anti-ai-writing.md"] |
| triggers | ["/atom:fix-text","修正文本","去AI腔","替换禁用词","标点修正"] |
fix-text
功能
修正文本质量问题,包括AI腔句式、禁用词替换和标点规范化。
双模执行
- 独立模式:输入为空时,读取
正文/ 目录下最新章节
- 编排器模式:输入不为空时,以传入的 {chapter_file} 为准
执行步骤
- 读取 {chapter_file}
- 根据 {fix_type} 执行修正:
ai:消除AI腔模板句式("不是A而是B"、万能状语、比喻三连、排比句式)
banned:替换禁用词为推荐表达(参考 banned-words.md)
punctuation:运行 node _shared/scripts/punctuation-normalize.js {chapter_file}
all:依次执行三项修正
- 输出修正后文本
AI腔修正规则
- "不是A,而是B" → 直接描写B或用更自然的转折
- "带着X"万能状语 → 拆短句或换具体动作
- "仿佛/宛如/恰似"比喻 → 删除或改为白描
- 排比句式 → 打散,变化句长
禁用词修正规则
- Level1 禁用词(31词)→ 命中即替换
- Level2 禁用词(18词)→ 命中需审查后替换
输出格式
文本修正报告
- 文件:{chapter_file}
- AI腔修正:N 处
- 禁用词替换:N 处
- 标点修正:N 处