Course quality editor
Skill Ivor-NCUT/course-producer/skills/course-quality-editor
面向 Codex Goal 长时运行的课程制作 Agent Skill:课程定位、专家访谈、材料加工、逐字稿、审校与课程海报。
npx -y skills add Ivor-NCUT/course-producer --skill course-quality-editorAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 20 days oldThe repository was created 20 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- 0 stars0 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.
What its author says it does
Copied from the file, not written here
对整门课程和单章执行认知交付、模块顺序、逻辑连续、信息密度、口播节奏、事实证据与 AI 写作坏味道审校,生成 review.jsonl 和保留讲师立场的修订稿。Use when the user says 课程审校、检查课程逻辑、去 AI 味、检查逐字稿、精修课程、检查事实证据,或 Course Producer 进入 quality_review 阶段。
SKILL.md
3.7 KB, ~1.2k tokens by cl100k_base, as published. Nobody here has run it
课程质量审校
围绕已有稿件审校和修订,不另起炉灶重写整门课。开始前读取课程定位、蓝图、知识卡、全部章节,以及 知识库/方法论/课程审校.md。
审校顺序
- 运行
node tools/course-lint.mjs <lessons-dir> --output <review.jsonl>,得到可复现的禁用句式、模板转场、字幕式断行和证据提示。 - 做课程级检查:模块是否按学习成果排序,第一、第二模块能否独立交付结果,章节是否螺旋复用,承诺是否越过定位边界。
- 做章节级检查:教学任务、认知起终点、挑战情境、段落到段落的因果/动作连接、信息密度、术语解释和前后桥接。
- 核查事实证据:数字、案例、背书、高风险承诺是否对应
knowledge_card_id与来源;找不到时标为补材料,不能润色成确定事实。 - 对每个问题选择动作:
auto_fix:不改变观点的措辞、重复、缺桥接、段落节奏与明确格式问题;needs_evidence:缺数据、案例、授权或原始来源;needs_decision:课程承诺、品牌立场、目标人群、价格、公开风险或证据冲突需要用户判断。
- 先复制原稿到
.course-producer/artifacts/revised-lessons/再应用auto_fix。不覆盖已确认原稿,除非用户明确要求原位修改。 - 把所有发现逐行写入
.course-producer/artifacts/review.jsonl;修复后更新status与revision_locator,保留问题原始证据。 - 重新运行确定性扫描并做课程级复核。高风险问题未解决时不能把 quality_review 标记完成。
review.jsonl 契约
{"review_id":"rev-001","scope":"lesson","artifact":"lessons/01.md","locator":"## 方法","category":"logic_gap","severity":"high","evidence":"上一段产物没有进入本段动作","action":"auto_fix","status":"fixed","suggestion":"补充产物到动作的桥接","revision_locator":"revised-lessons/01.md#方法"}
category 可使用 course_order、logic_gap、repetition、density、spoken_flow、ai_smell、unexplained_term、unsupported_claim、position_conflict。不要输出主观“AI 概率”。
保真边界
- 可以改结构、顺序、句群、转场、重复和术语解释。
- 不改变讲师观点、战略立场、案例因果、数字口径和课程承诺。
- 对观点有疑问时写
needs_decision,不要用审校者立场替换。 - 逐字原话只在确认是引用错误时改;普通口语不因“不够精致”被抹平。
完成检查
- 课程级与章节级审校都已执行。
- review 每条有位置、具体证据、严重度、动作和状态。
- 自动修复后的稿件可回放,原稿仍可恢复。
- 数字、案例、背书和高风险承诺有来源或明确待补。
- 没有用概率分数代替具体文本证据。
- 修订没有改变讲师观点和战略立场。
完成后可进入 course-lark-delivery;用户对修订结果反馈时按审美对齐入口处理。
Feedback Learning
用户对修订稿给出反馈、亲自改稿、选择版本或明确满意时,先判断是否只是当前作品修改。用户确认要沉淀后,调用 course-aesthetic-alignment 记录 edit trace、正反例或 preference pair;单次满意不能直接更新本 Skill 的审校规则。
What ships with it: 1 file
273 B alongside SKILL.md
agents/
- openai.yaml273 B