ljg-read
Thoroughly explain user-provided text using the "bottom-up structure" from "A View of Learning": bridge concrete scenarios to concepts, classification criteria, and conditional relationships, enabling readers to grasp the original meaning and transfer their judgment; displayed directly in the conversation by default. USE WHEN the user says "read along", "read with me", "read this", "explain this passage thoroughly", or provides text seeking an accessible yet in-depth explanation. NOT FOR simple translation, full-book breakdowns, or merely drawing concept maps.
1How It Works & Mechanism
Use the "bottom-up structure" from "A View of Learning": place concrete objects and scenarios at the bottom layer, concepts and general relationships at the top layer, prioritizing practical problems.
**下层是实际问题，上层是理解问题的工具。**
Discriminative models bridge objects and concepts: when explaining a concept, enable the reader to see its referent and classification criteria.
**判别模型接通对象与概念。** 它回答「眼前这个情况，凭什么算这一类？」
Associative models bridge relationships between concepts: explain correct classification and relational validity separately.
**联结模型接通概念之间的关系。** 它回答「知道了这一类情况，在什么条件下能推出什么？」
Validation: restating explained examples is insufficient to prove transfer; present new scenarios for reader judgment.
**材料和验证承担不同作用。** 具体例子帮助建构，抽象表述帮助整理，未见情况用来检验迁移。
Source text and scope: read the actual content first; if the full text cannot be retrieved, state the gap rather than hallucinating text.
拿不到正文时说明缺口，不根据标题或摘要补写全文。
Six bridging requirements: make things visible, concepts distinguishable, relationships traceable, two layers aligned, boundaries shift judgment, and transfer verifiable.
- **让事情可见。** 优先使用原文例子；原文太抽象时补一个贴近日常的教学设例，并就地标明。
Displayed only in the dialogue by default, without automatically generating files, memories, or learning logs.
**伴读结果默认只展示在对话中。** 不自动创建或更新笔记、Org 文件、对话记录、学习档案、记忆或其他产物，不返回虚构的文件路径。
Optional interactions: fast-forward / expand / wait a moment / interactive self-test; asking questions does not gate delivery.
- **快进**：压缩铺陈与次要解释，保留主判断、关键对应与适用条件。
Treat argumentative texts and narrative/lyrical texts separately; distinguish character motives shown in the text from speculation.
论证文本重在概念、依据与推理；叙事或抒情文本重在具体处境、人物变化、意象与表达效果。
Finished conversation format: defaults to coherent Chinese Markdown paragraphs, concluding with a single sentence capturing the key takeaway.
默认中文 Markdown，以连贯段落讲解。短文本不硬分章；长文本用贴合内容的小标题。
2Core Capabilities
4Recommended Governance & Evidence
- SKILL.md: Pure prompt instructions and rule text; no script execution or external impact
5Second-Pass Verification
- [ok] 实现原理＝纯 prompt 编排，无自带可执行代码 — 目录仅 SKILL.md（8946 bytes）；风险扫描 URL/凭据/外部程序/写盘/网络/危险旗标 全线 0 命中，token 扫描同样无 fetch/curl/child_process/process.env。
- [ok] 「下上结构」的两层定义前后一致 — 题头与正文都写下层＝实际问题/具体对象、上层＝通用关系/工具，并强调两层相对可互换、不能把下固定成物质上固定成语言，无矛盾。
- [ok] 方法论出处属实 — 明确署名依据于建国《学习观：从感觉懂了到真正学会》的下上结构、判别模型、联结模型与学习验证论述，并自称讲解安排是本技能的应用设计。
- [ok] 默认零产物、不返回虚构路径的约束明确 — 正文加粗强调『伴读结果默认只展示在对话中』，列明不创建笔记/Org/对话记录/学习档案/记忆，原文保持只读，仅用户明确要求才保存。
- [ok] 外部依赖为零 — 风险扫描七类全部 0 命中；正文只说 URL 用『当前可用的网页获取能力』，未点名任何具体工具或域，故 external_deps 与 network_calls 均空。
- [ok] frontmatter 自述版本与 git commit 不是同一体系 — metadata.version 为 "2.0.0"（skill 自述版本），audit.skill_commit 为 git commit 9e75497…；两者不混用，结论只对 pin commit 有效。
- [ok] meta 直采项无可定位 per-skill 值 — §0 未提供 stars/last_push，installs 无 per-skill 可定位值，按契约均保持 null。
6Verdict & Scenarios
df27a93b98396555…