ljg-teach
"Teaches concepts, principles, or tests perspectives through experience rather than definitions: first presents a scenario matched to the learner's cognitive level, prompts a choice, reveals cognitive gaps through a fair surprise, structures the insight before naming it, and repeatedly retests with transfer scenarios in a loop. USE WHEN the user invokes '/ljg-teach', says 'teach me through scenarios', 'don't tell me the answer first', 'help me learn through cases', or wants to understand a concept or viewpoint through choices and feedback. NOT FOR direct definition lookups (use ljg-is/ljg-word), eight-dimensional concept dissection (use ljg-learn), or companion reading (use ljg-read)."
1How It Works & Mechanism
The core is an experience-first six-step learning loop: Scenario → Choice → Surprise → Structure → Naming → Transfer Scenario.
局面 -> 选择 -> 意外 -> 结构 -> 命名 -> 迁移局面
Workflow Routing routes the next step based on the learner's state; when introducing a new concept, enter 'Scenario' first, strictly forbidding explaining or naming the target concept.
进入「局面」；不要解释或命名目标概念
Scenario design requirements: The learner has a role, goal, constraints that genuinely affect choices, and sufficient information; the target concept's name/definition/terminology or answer outlines must not appear in the first round.
首轮不出现目标概念的名字、定义、术语暗示或答案轮廓。
Ask only one decisive trade-off judgment, permit answers outside given options and 'insufficient information', conclude with 'What would you choose? State your reason in one sentence' and stop to wait for the answer.
只提出一个必须取舍的判断。允许选项外回答，也允许「信息不足」
Surprises must be fair: do not penalize insufficient information by hiding facts, do not use brain teasers or rare exceptions to disguise gaps, and do not fabricate failures.
不捏造失败。如果学习者的选择在给定约束下成立，就承认成立
In the structuring phase, mirror the learner's old model first without prematurely injecting theory; explicitly mark speculations when evidence is insufficient.
根据学习者两次回答，先把他的旧模型镜像出来，不抢先塞入理论
Naming must strictly occur after choice, surprise, and structuring; introduce only one concept per round that most effectively compresses the surprise.
结构说清之后，才说：「你刚才遇到的这个结构，叫……」
Transfer scenario: Present a new scenario that is superficially different but structurally identical without hinting at applying the concept; correct answers with rationales still rooted in the old model do not count as mastery.
立即给出一个表面不同、深层结构相同的新局面，不提醒该套用刚命名的概念。
Interaction pacing: Advance only one cognitive breakpoint at a time, include only one load-bearing question per message, and strictly wait for the learner's response between choice and surprise.
一次只推进一个认知断点，一条消息只放一个承重问题。
Emphasize not turning viewpoint instruction into ideological conditioning: for value disagreements, only highlight costs, constraints, and predictive differences, preserving the learner's right to prioritize their own values.
不把观点教学做成立场驯化。
2Core Capabilities
4Recommended Governance & Evidence
- SKILL.md: Pure prompt instructions and rule text, without script execution or external impacts
5Second-Pass Verification
- [ok] 六步体验循环 — 『局面 -> 选择 -> 意外 -> 结构 -> 命名 -> 迁移局面』原文在。
- [ok] 状态路由 — 『进入「局面」；不要解释或命名目标概念』原文在。
- [ok] 局面首轮不泄露概念 — 『首轮不出现目标概念的名字、定义、术语暗示或答案轮廓。』原文在。
- [ok] 公平意外三条禁令 — 『不捏造失败。如果学习者的选择在给定约束下成立，就承认成立』等原文在。
- [ok] 命名后置 — 『结构说清之后，才说：「你刚才遇到的这个结构，叫……」』原文在。
- [ok] 迁移复测 — 『立即给出一个表面不同、深层结构相同的新局面』原文在。
- [ok] 无落盘 — 『没有要求时不写文件、不生成笔记、不布置课程』；扫描写盘零命中，一致。
- [ok] NOT FOR 点名他人 skill — description NOT FOR 列出 ljg-is/ljg-word/ljg-learn/ljg-read；按口径仅记入 reason，不改档位。
6Verdict & Scenarios
c681a74730bbaf9a…