Skip to content
Education
Skill

/lecture-style

Masterclass lecturing — one continuous explanation carried by slides, with dense pages, long narration that builds an argument across several real cases, and checkpoints that are rare and deliberate. Use when the ask is a lecture, a masterclass, a systematic walkthrough of a

From plugin
openmaic
43k25 skills

Claude Code

$ npx -y skills add thu-maic/openmaic --skill lecture-style --agent claude-code

SKILL.md

lecture-style.SKILL.md
name: lecture-style
title: "大师讲授"
description: Masterclass lecturing — one continuous explanation carried by slides, with dense pages, long narration that builds an argument across several real cases, and checkpoints that are rare and deliberate. Use when the ask is a lecture, a masterclass, a systematic walkthrough of a subject, 「大师课」「系统讲解」「讲透」, or any topic whose payoff is understanding rather than practice; also use when the user names this style. Not for hands-on training (that is `workshop-style`) and not for a course whose point is manipulating a mechanism (that is `deep-interactive`).

Masterclass lecture design

You are designing a **lecture**. One voice carries the learner from not knowing the subject to being able to explain it, and the pages exist to support that explanation rather than to interrupt it. Density is a feature here: a lecture page that says one thing is a wasted page.

`stage-design` still governs how the stage is built — outline, roster, one page per `generate_scene`, audio before done. This skill governs what the stage contains and how it sounds.

The shape

  • **Scene 1 is a `slide`** that opens the lecture properly — the question the

subject answers, why it is worth an hour, and the route the lecture takes through it. Not a definitions page and not a table of contents.

  • **The body is `slide` pages, in arcs.** An arc is three to five pages that

carry one idea from statement to consequence — the claim, the mechanism underneath it, then real cases that test it. Arcs follow each other; the lecture never restarts.

  • **Checkpoints are rare.** One `quiz` roughly every four to five pages, placed

at the seam where one arc closes and the next begins, testing whether the learner can apply the arc's idea rather than recall its words. Two quizzes in six pages is not a lecture, it is a test with slides in between.

  • **At most one `interactive` page**, at the single most mechanism-heavy concept

in the whole lecture — the one place where telling genuinely fails and the learner has to see the thing move. If the subject has two such concepts, a second is defensible; a third means the request was never a lecture and `deep-interactive` fits it better.

  • **Close on the argument**, not on a bullet recap: what the lecture

established, what it deliberately left open, where a curious learner goes next.

Page density

Each body page carries **four to six `keyPoints`** and they are propositions, not labels — 「利率上升先压估值再压盈利」rather than 「利率影响」. A page holds one claim plus the evidence that makes it stick: a definition and its boundary, a mechanism and a worked case, a comparison and the criterion that separates the two sides.

Name a real case per arc — a company, a study, a historical episode, a piece of code — and let it recur. A lecture that cycles through a fresh unexplained example every page teaches nothing about any of them.

Narration is the lecture

This is where the style is heard, so it is the part to get right.

  • Every body page gets **a long, continuous stretch of teacher speech** — four

to six sentences per line rather than a caption, and several lines to a page. The learner should be able to close their eyes and still follow.

  • Narration **moves forward**. Each line picks up the last one's final idea:

state it, explain the mechanism, then walk a concrete case through that mechanism until the conclusion is unavoidable.

  • The register is an authority thinking aloud in front of a room — full

sentences, confident, unhurried, occasionally naming what is hard about the idea before resolving it.

  • The assistant speaks **rarely**, and only to ask the question the room is

already thinking, which the teacher then answers at length. One such exchange per arc, at most.

  • Forbidden in this style: 「大家想一想」-style prompts with nothing behind

them, one-line captions that restate the slide's title, and cheerleading (「太棒了」「让我们开始吧」).

Examples of the right length and motion:

  • Good: 「我们先把定义钉住——所谓久期,不是债券还剩多少年,而是价格对利率的

敏感度……接下来看 2022 年的例子,同样的加息幅度,为什么二十年期的跌幅是五年期 的四倍多。」

  • Bad: 「这一页讲久期。久期很重要。下面我们来看例子。」

The roster

The runtime allows **exactly one teacher**, and in this style that teacher is the lecturer — the whole course is their voice. Write the roster with `set_roster`: a senior domain authority with a calm, measured delivery, plus one assistant whose entire job is to raise the sharp question at an arc boundary. Keep the roster small; a crowded classroom fragments a lecture. Put the lecturing manner in the persona text and the pacing in `voiceDesign.delivery` — 「calm measured authoritative, unhurried」 rather than 「lively energetic」.

Carrying the style into the generators

There is no outline generator: you plan the course in the conversation, then call `create_stage` and one `generate_scene` per page with an explicit brief. Page content and narration are written by separate calls that see only the page's brief, the page content and the roster. So the style has to be carried there explicitly:

  • put the arc structure and the density expectation into each page's

`generate_scene` `brief`;

  • put the lecturing manner into each persona of `set_roster`;
  • pass the narration directive for each page in `generate_scene.materialFacts`

— e.g. 「旁白为连续讲述,四到六句一段,先定义再机制再案例」「本页复用前一页 的同一家公司作为案例」. Naming this skill in a prompt does nothing; only the text you place in those fields is seen.

Scene naming

Titles are the lecture's own steps, in the language of the subject.

  • Good: 「久期到底在度量什么」「同样加息,为什么长债跌得更狠」「三个反例,和它们

的共同点」

  • Bad: 「概念介绍」「案例分析」「本课总结」

If the request is not a lecture

If the user wants to practise a skill, produce something of their own, or work through exercises, say so in one sentence and use `workshop-style` instead. Wrapping a practice request in a lecture produces a course the learner watches and cannot use.

Read more
@thu-maic
Ships withopenmaic

Open Multi-Agent Interactive Classroom — Get an immersive, multi-agent learning experience in just one click

Stars on GitHub

Momentum

stars in total
43.2k
Star history of thu-maic/openmaic on GitHub, a chart by star-history.com

Other skills on openmaic.

build-personal-skill

build-personal-skill

Evidence-based creation of a reusable personal course-making Skill from the user's own classroom and chat history. Use when the…

@thu-maic@thu-maicInstall Skill
curriculum-planner

curriculum-planner

Multi-stage series — a request for several classrooms that belong together, like 「7 天学 Python」, a four-week onboarding track, a…

@thu-maic@thu-maicInstall Skill
deep-interactive

deep-interactive

Deep Interactive courses. Plans the course so the learner manipulates something on most pages — simulations, diagrams they…

@thu-maic@thu-maicInstall Skill
deep-research

deep-research

Courses whose content rests on current, external or real-world facts that must be verified against live sources before being…

@thu-maic@thu-maicInstall Skill
fact-check

fact-check

Improve factual reliability while creating or reviewing a course or supplied content. Use when the user asks to fact-check,…

@thu-maic@thu-maicInstall Skill
feynman-learning

feynman-learning

Turn any concept, lesson, slide deck, or source material into a Feynman learning cycle in which learners explain first, expose…

@thu-maic@thu-maicInstall Skill

More skills from other plugins.

lesson-quiz

lesson-quiz

Test a learner on a single Claude Code tutorial lesson (01-10) with 10 questions, scoring answers and flagging weak spots. Use…

@luongnv89@luongnv89Install Skill

Questions about lecture-style.

Which coding agents can use lecture-style?
lecture-style installs in Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, OpenCode and Windsurf with the skills CLI, using each agent's flag shown on this page. Flowy's auto-invocation, which fires a skill when your prompt matches, runs in Claude Code only.