Stage Design logo

Stage Design

OrganizationPopular
THU-MAIC
stage-design

The baseline method for building ONE stage — a single classroom — with this runtime. Covers how the plan is settled in conversation, the order the generation tools are called in, how the classroom roster is written, when narration has to be re-synthesized, and what must be true before a stage is called done. Load this before generating or rebuilding any stage from scratch, whatever its subject; it is the floor the topic skills build on, not an alternative to them.

Overview

PublisherTHU-MAIC
RepositoryOpenMAIC
Skill namestage-design
Stars
37.6K
Forks
5.9K
Bundled files
Instructions only
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by THU-MAIC on GitHub. Read the source before you install it.

Installation

Install the Stage Design AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/THU-MAIC/OpenMAIC.git /tmp/OpenMAIC
mkdir -p .claude/skills
cp -r /tmp/OpenMAIC/skills/agent-runtime/stage-design .claude/skills/stage-design
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Stage Design in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Stage Design on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Stage Design is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Building one stage

A stage is one classroom: one settled plan, one roster, one deck of pages. This is how a stage is built here, whatever its subject.

A topic skill (deep-interactive, vocational, deep-research, teacher-style-clone, …) may shape what the stage contains — its structure, its page types, how scenes are named — and where one is active its instructions win on those questions. The sequence below is not one of those questions: it is the same under every topic skill. Read this one first, then the topic skill.

Nothing here applies to a stage that already has pages. A stage with persisted pages is edited, not built (see pro-editing).

The build sequence

Build a NEW stage only when there are no persisted pages yet. Then, in this order:

  1. Plan the stage in the conversation. Propose the page plan yourself — each page with its title, type (slide / quiz / interactive / pbl) and brief (this page's teaching intent and content outline, the same information a planned outline slot used to carry) — in page order. Settle the plan with the user through ask_user before creating anything: there is no outline tool to plan with, so the plan is your words and their sign-off. A topic skill's structural constraints are a diagnostic the runtime checks against the real stage after each page; you decide how to act on it.
  2. create_stage with the settled course title (pass the series folder's folderId when the stage is part of a series, so it is filed in the same call). From here on, pass the returned stageId to every stage tool.
  3. set_roster, before any page is generated. Write the classroom the user settled: exactly one teacher, at least two agents in total, each with a concrete persona (2-3 sentences on personality and teaching/learning style, in the stage language) and a voice — call list_voices first and bind one of the returned providerId::voiceId pairs, or the exact pair register_voice returned for a cloned voice; omit the binding when no usable voice exists. Every page's content and narration is written for the roster that exists when the page is generated, so a roster settled late is a roster half the stage never saw.
  4. generate_scene for EACH settled page, in ascending order — one page per call, never more. Pass the page's settled title, type and brief (plus materialFacts when the page's content comes from attached material); no planned outline is needed. A pbl page is generated the same way and lands as a full PBL project (projectV2) from its brief; a failed generation reports an error and writes nothing. Each call is the durable checkpoint for its page: the page survives an interruption only once the call has returned.
  5. list_scenes to verify every page you settled is actually persisted.

None of the planning steps is a place to stop: a turn ends when a page actually landed, when ask_user is waiting on the user, or when the stage is done by the check below. Reporting what you have planned, imported or confirmed leaves the user holding an empty deck.

When generate_scene fails

Track generation attempts by stageId + order. Changing the title or brief, or generating another page in between, does not reset the attempt.

  • For prompt-unavailable or another deterministic failure, do not repeat the same call.
  • For invalid-model-output, change the brief without changing the settled teaching intent, then retry once. Retry once for a clearly transient provider failure too.
  • If that retry fails, do not call generate_scene again for that target during this run. Leave the page unwritten, continue with later pages in the same stage, and name the gap when wrapping up. Do not call ask_user after every failed page or silently drop the page from the settled plan.

Use list_scenes to reconcile persisted pages with the settled plan. A stage with a missing page can be wrapped up after later pages are built, but it is not done; name each gap and its latest failure instead of claiming completion.

Narration audio follows narration text

When a page's narration is written or reworded with patch_stage on an /actions/... JSON Pointer, call generate_tts on that page. A reworded speech line has no audio until the page is re-synthesized, so narration text that was changed and not re-synthesized ships as a silent page.

Before you call the stage done

Every settled page has to be persisted — a stage is done when the deck is complete, not when it is planned or half built.

Every speech action must have audio. read_stage each page at path:/scenes/<order|id> with detail:"source", check the audioId field on its speech actions, and call generate_tts on any page that is missing it. Read what those actions say while you are there: narration that talks about material this stage does not teach is a defect, whatever its audio status. Only then is the stage finished.

When every settled page is persisted, say so to the user in one or two sentences.

Frequently asked questions

What does the Stage Design AI skill do?

The baseline method for building ONE stage — a single classroom — with this runtime. Covers how the plan is settled in conversation, the order the generation tools are called in, how the classroom roster is written, when narration has to be re-synthesized, and what must be true before a stage is called done. Load this before generating or rebuilding any stage from scratch, whatever its subject; it is the floor the topic skills build on, not an alternative to them.

Why use Stage Design on TypingMind?

Because you install it once and use it with any model. Stage Design is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Stage Design in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/THU-MAIC/OpenMAIC/tree/main/skills/agent-runtime/stage-design. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Stage Design?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Stage Design?

As many as you like. As long as a model supports skills, you can use Stage Design with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Stage Design AI skill free?

Yes. It is published on GitHub by THU-MAIC under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇