Labs often need supporting context — background on a product, a refresher on a concept, a check that the learner has understood — but writing all of it into the instructions produces a wall of text most learners skip. Menu Assistant turns a word or phrase in your instructions into a link that generates that supporting content on demand, so it's there for the learners who want it and out of the way for those who don't. This article covers how to configure Menu Assistant, how to write a topic and system prompt that produce consistent responses, and when to choose it over Chat Assistant.
Best Practices for Menu Assistant
The Topic and the System Prompt do almost all the work. The Topic is both the link text the learner sees and the subject sent to the model, so it has two jobs at once; the System Prompt is what stops four different learners getting four differently-shaped answers.
Write the Topic as the words you want in the sentence. It appears inline in your instructions, so it should read naturally where it sits —
ai-menu[Azure Storage Accounts]inside a sentence about storage, not a bare keyword dropped into a paragraph.Start the Topic with a verb when you expect learners to use Show. "Create an Azure Storage Account" produces a usable set of steps; "Azure Storage Accounts" produces an explanation with steps bolted on.
Use the System Prompt to fix the shape of the response. Length, depth, format and scope all belong here. Without it, response quality depends entirely on how the model interprets a two-word topic.
Limit Options to what the topic can actually support. A conceptual topic rarely produces a good Show response, and a procedural one rarely produces a good Quiz. Offering all four on every menu invites learners into the weakest one.
Match Response Phrasing to your audience, not to the topic. A 100-level lab and a 400-level lab can share a topic and still need different phrasing.
Don't rely on Quiz for assessment. Menu quizzes are practice, not scoring — they aren't recorded against the learner and the interactive formatting isn't guaranteed. Use a Question or Automated activity when the result needs to count.
Check the token budget before you publish. Every generated response consumes AI tokens against the Max AI Tokens allowance on the Lab Profile. A lab with menus on every page can exhaust it before the learner reaches the end.
Generate every option once before publishing. Launch the lab, click through Teach, Show, Quiz and Resources on each menu, and read what comes back. This is the only way to catch a topic that reads well but prompts badly.
If what you need is a back-and-forth rather than a canned response — a learner describing a problem and working towards the cause — use Chat Assistant instead. Menu Assistant is the better choice when you know in advance which four kinds of help a learner might want, and it's cheaper and more predictable for it.
Who Uses Menu Assistant, and When
Lab Developers configure Menu Assistant in Skillable Studio while authoring a lab. Learners use it inside the lab client during an active lab session.
Reach for it when supporting content is useful but shouldn't be compulsory reading:
Product or concept background. A lab that assumes familiarity with a platform can offer that familiarity on demand rather than spending a page establishing it.
Mixed-audience labs. Learners who need the explanation get it; learners who don't, don't have to scroll past it.
Self-checks between sections. A Quiz menu lets learners test their own understanding before moving on, without you authoring and scoring question activities.
Further reading. Resources gives learners somewhere to go next without you maintaining a link list that rots.
Content you'd otherwise write four times. One menu on a recurring concept replaces the same explanation repeated across several labs.
Avoid it where the content must be exact, reviewable or identical for every learner — compliance material, anything a customer audits, or steps a learner will be scored against. Generated responses vary between learners by design.
How Menu Assistant Works in Skillable Studio
Menu Assistant is one of the learner-led tools in the Skillable AI suite. You insert it as an element wrapped around a topic in your instructions. At lab launch the topic renders as a link; selecting it opens a menu of up to four options, and the learner's choice generates a response that appears in a collapsible box below the current paragraph.
Insert Menu Assistant from the Instructions Editor
From Skillable Studio, open your lab's Instructions Editor.
Place your cursor where the topic should appear in your instructions.
On the ribbon bar, select AI > Menu. The Insert Menu dialog opens.
Complete the configuration fields:
Setting | Description |
|---|---|
Topic | The word or phrase learners select in the instructions, and the subject sent to the model. Required. |
Model | The model that generates responses. This field only appears if your organization has more than one model available; organizations using the default Skillable model have nothing to select. |
System Prompt | Instructions and constraints applied to every response generated from this menu. Use it to set depth, length, format and scope. |
Options | Which of the four menu options learners can choose from. All is selected by default; clear it to pick individual options. |
Response Phrasing | How simply or technically responses are phrased: Simple, Default or Complex. |
# Quiz questions | How many questions a Quiz response generates. Only applies if Quiz is among your selected Options. |
Select Insert. The topic and its menu are added to your instructions.
Existing Menu Assistant elements can be reopened and edited by double-clicking them in the visual editor.
The four menu options
Menu Option | Description |
|---|---|
Teach | Explains the topic to the learner. |
Show | Produces a list of steps for performing the topic. Works best when the Topic starts with a verb, for example "Create an Azure Storage Account". |
Quiz | Produces multiple-choice questions on the topic. Where the model returns a parseable format, questions render interactively — a correct selection returns "Correct" and an incorrect one "Sorry, try again." Occasionally a response can't be formatted and appears as plain text instead. |
Resources | Produces a list of links to web resources about the topic. |
Underlying syntax
The dialog writes an IDLX element into the instruction markdown, and you can write or edit that element by hand in code mode. The basic form wraps the topic in brackets after ai-menu:
ai-menu[Azure Storage Accounts]
Parameters are added in braces after the topic. Everything between the braces is JSON, so keys and string values must be in double quotes, booleans written as true or false, and numbers written in full:
ai-menu[Create an Azure Storage Account]{"prompt":"Assume the learner is new to Azure. Keep steps under ten and reference the portal, not the CLI.","options":"teach,show","user_level":"simple"}
Two things to expect when you switch to code mode after using the dialog. It writes a model_id recording the model you selected — this value is specific to your organization's configuration, so a snippet copied into a different organization will point at the wrong model or none at all. Remove it if you're sharing syntax between organizations. It also only writes values that differ from the defaults, so a menu left on All options and Default phrasing produces no options or user_level entry.
Working in code mode gives you access to parameters the dialog doesn't expose. Two are worth knowing about.
Per-option prompts. teach_prompt, show_prompt, quiz_prompt and resources_prompt let you shape each response type separately, instead of applying one System Prompt to all four:
ai-menu[Azure Storage Accounts]{"teach_prompt":"Explain the storage account as a container for services. Avoid CLI syntax.","quiz_prompt":"Ask about redundancy options and access tiers only.","options":"teach,quiz"}
Custom options. custom_options replaces or supplements the four built-in menu items with questions you write yourself, and custom_prompts supplies the prompt sent for each one. The two arrays pair by position — the third custom option uses the third custom prompt. To show only your own options, set options to an empty string and use_default_prompts to false:
ai-menu[ETL]{"options":"", "use_default_prompts":false, "custom_options":["What is ETL?", "Detailed explanation of ETL"], "custom_prompts":["Define what is the ETL process", "Detailed explanation of the Extract Transform and Load process for people with experience"]}
Custom options must be entered by hand in code mode; the Insert Menu dialog has no field for them. For guidance on writing prompts, see Effective AI Prompting.
How responses behave at runtime
Responses appear below the current paragraph in a collapsible box, so learners can tuck them away once read.
Responses carry like and dislike controls. Learners can rate what they were given.
Each option generates independently. Selecting Teach and then Quiz produces two separate responses, each consuming tokens.
Usage counts towards Max AI Tokens per Lab Instance. When the ceiling is reached, AI features stop responding and the message configured on the Lab Profile is displayed instead.
Worked example
You're building a lab on Dataiku, and learners need enough background on the platform to follow the exercises — but that background isn't the point of the lab, and a page of it up front would get skipped. A menu on the product name gives learners the introduction only if they want it.
Enter Dataiku as the Topic, and in the System Prompt describe what the response should cover: an overview and introduction to the Dataiku platform.
The resulting markdown is:
ai-menu[Dataiku]{"prompt":"Provide the user with overview/introduction information about the Dataiku platform"}
In the instructions, "Dataiku" now renders as a link wherever it appears in your text.
A learner who selects Teach me gets a generated introduction to the platform, in a collapsible box below the paragraph they were reading.
Next Steps
Before publishing a lab that uses Menu Assistant, set the AI token ceiling and the message learners see when it's reached. Both are configured on the lab's Lab Profile under Max AI Tokens per Lab Instance. Menus are cheap individually but add up across a long lab, so this decides whether later menus still work for a learner who used the earlier ones.
Related Topics
Chat Assistant — open conversation scoped to a topic, for when learners need to ask their own questions rather than choose from four options.
Dynamic Teaching — content generated at launch, for when it should be there before the learner asks.
Skillable AI — how Menu Assistant fits alongside the rest of the suite, and the difference between Skillable pre-integrated and BYO models.
Practice Assistant — for learner-generated hands-on scenarios rather than explanatory content.
Instructions Editor — the full authoring surface, including code mode for editing Menu Assistant parameters directly.
Effective AI Prompting — principles for writing System Prompts that produce consistent responses.
Automating Activities — for scored assessment, where a Quiz menu isn't a substitute.