メニュー
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).
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
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 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 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.
| name | project-onboarding |
| description | 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). |
Orient yourself in a learning project workspace. Read-only. Do not modify files.
Host prompt will include the project path. If $ARGUMENTS.project_path is given, use it; otherwise the agent SDK's cwd is the project root.
Use Read and Glob tools only. No Write, no Edit, no Bash.
{project_path}/.learning/project.json. Expect fields: name, created, chapters[], total_duration. If missing, the project hasn't been initialized — tell the user and stop.{project_path}/*.md to find chapter files (zero-padded, e.g. 00-*.md, 01-*.md). Sort by name to get generation order.*.concepts.json to extract top concepts. If none, the project is fresh.Return a single markdown text response (the host displays it as the agent's first message):
## 项目摘要
**主题**: {project.name}
**进度**: 已完成 {N}/{M} 章 ({percent}%)
**当前焦点**: {latest chapter title or "尚未开始"}
**已学概念**: {top 3-5 from concepts.json, comma-separated; "无" if fresh}
**接下来**: {one-line suggestion: 完成第 N+1 章 / 复习 / 等等}
Keep it under 200 words. No code blocks, no Mermaid.
project.json missing → respond 项目尚未初始化 and exit*.md files yet → say 已规划 {M} 章但尚未生成任何内容