菜单
Generate quiz questions from saved Cornell explanation cues (concept + Q&A history). Use when the host invokes "请用 extra-quiz-generation skill 根据以下概念列表生成附加题" with a JSON list of concepts and their Q&A histories. Writes the result to disk using the Write tool.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Generate a structured reading guide for an academic paper markdown file. Use when the host invokes "请用 typora-paper-reader skill 为以下论文生成导读" with a paper file path and user persona. Reads the paper, identifies key points based on the persona, writes a JSON guide to disk.
Generate review cards (quiz questions + key points) for each concept in a completed learning chapter. Use when the host prompts with chapter content and concept list. Outputs structured JSON only — does not write files.
Conduct a Socratic dialogue to deepen understanding of a concept cluster. Use when the host prompts "请使用 typora-socratic-review skill 进行苏格拉底复习". Returns one Socratic question per turn in markdown; end session with `[SESSION_END]` marker when the user demonstrates mastery or asks to stop.
Explain selected text from a learning chapter using analogies and plain language, then suggest follow-up questions for deeper understanding. Use when the host invokes "请用 explanation skill 解释" with a term, context, and optional previous Q&A history. Writes the result to disk using the Write tool.
Generate a complete learning chapter (Markdown + quiz.json + concepts.json) for one slot in a learning project outline. Use when the host prompts "请使用 chapter-generation skill 生成第 N 章" with chapter index/title/duration/concepts. Writes three files via the Write tool. Skill is read-only and self-contained — does not call other skills.
Read-only orientation for a new agent session in a learning project. Use when the host prompts "请使用 project-onboarding skill 了解这个项目". Read project state, summarize in markdown, do NOT modify any files. Run once per session (typically at init).
| name | extra-quiz-generation |
| description | Generate quiz questions from saved Cornell explanation cues (concept + Q&A history). Use when the host invokes "请用 extra-quiz-generation skill 根据以下概念列表生成附加题" with a JSON list of concepts and their Q&A histories. Writes the result to disk using the Write tool. |
Generate single-choice quiz questions from the user's saved Cornell explanation cues. Each cue represents a concept the user highlighted and asked about while reading.
The host's prompt will contain a JSON list of concepts:
[
{
"concept": "位置编码",
"qa_history": [
{"q": "位置编码", "a": "位置编码给每个 token 加一个位置向量..."},
{"q": "为什么用正弦余弦？", "a": "因为正弦余弦函数具有周期性和连续性..."}
]
}
]
Each entry has:
concept — the user's highlighted term (concept name)qa_history — Q&A pairs starting with the initial explanation (Q0/A0), then follow-up Q&As用 Write 工具将以下 JSON 数组写入 prompt 中指定的 output_file 路径：
[
{
"id": "extra_1",
"qtype": "single",
"question": "关于「位置编码」，以下哪项描述最准确？",
"options": [
{"label": "A", "text": "位置编码给每个 token 加一个位置向量，让 Transformer 能区分序列顺序"},
{"label": "B", "text": "这是一个合理的错误选项（基于常见误解）"},
{"label": "C", "text": "另一个合理的错误选项"},
{"label": "D", "text": "第三个合理的错误选项"}
],
"correct": "A",
"weak_concepts": ["位置编码"]
}
]
Based on Q&A history for each concept: Use the actual explanation content from qa_history to construct Option A (the correct answer). The initial explanation (Q0/A0) and follow-up Q&As provide rich material.
Distractor quality: Options B/C/D must be plausible but incorrect. Draw from:
One question per concept: Generate exactly one question for each concept in the input list. Do not skip any.
Question phrasing: The question stem should be specific to the concept. Vary the phrasing:
题型: Only "qtype": "single" (single choice). correct is a single string like "A".
weak_concepts: Always include the concept name.
id format: extra_1, extra_2, etc. — sequential starting from 1.