| name | educates-workshop-authoring |
| description | Comprehensive guide for creating and configuring workshops for the Educates interactive training platform. Includes steps for creating workshops from scratch, configuring workshop definitions and content and writing workshop instructions. Use this skill when creating Educates workshops, configuring workshop settings or writing workshop content and instructions.
|
Educates Workshop Authoring Skill
This skill provides guidance for creating interactive workshops for the Educates training platform.
Initial Workshop Creation
When the user asks to create a workshop, follow these steps:
1. Gather Workshop Details
Collect the following information from the user or infer from context:
- Title: A short, human-readable title for the workshop
- Description: A one to two sentence description of what the workshop covers
- Name: A machine-readable identifier (lowercase, dashes allowed, max 25 characters, recommended
lab- prefix)
If the user provides a topic but not explicit values, propose reasonable defaults and confirm before proceeding.
2. Determine Workshop Location
Ask the user to choose one of:
- Use the current directory as the workshop root
- Create a new subdirectory using the workshop name
Deriving the workshop name from the directory:
If the user chooses to use the current working directory:
- Check if the directory name satisfies the naming convention (lowercase, dashes allowed, max 25 characters). Note: the
lab- prefix is recommended but not required when using the directory name.
- If it does, use the directory name as the workshop name
- If it does not satisfy the requirements, infer an appropriate name from the workshop topic or title, but confirm with the user before proceeding (unless they already explicitly provided a name)
3. Determine Workshop Requirements
The terminal application is always enabled by default. Always include it explicitly with enabled: true and layout: split for clarity.
Ask the user about additional session applications based on the workshop's technical requirements:
- Editor: Will users need to edit or view files in the browser?
- Kubernetes access: Will users run kubectl commands or interact with the Kubernetes API?
- Kubernetes console: Would a visual cluster view help users?
- Docker: Will users build or run containers?
- Image registry: Will users push container images?
- Virtual cluster: Does the workshop need cluster-admin operations?
- Git server: Will users need a local Git repository (e.g., for CI/CD pipelines)?
- Slides: Does the workshop include a presentation alongside the instructions?
Infer sensible defaults from the workshop topic. For example, a Kubernetes workshop likely needs editor, Kubernetes access, and console enabled.
4. Create Workshop Structure
Create the following directory structure in the chosen location:
<workshop-root>/
├── README.md # Workshop title and description
├── exercises/
│ └── README.md # Placeholder to ensure directory is preserved
├── resources/
│ └── workshop.yaml # Educates Workshop definition
└── workshop/
└── content/ # Workshop instruction pages (Markdown)
The exercises Directory
Always create an exercises/ directory as part of the initial workshop structure. This directory serves as the working area for the workshop user — place any files they will need during the workshop here, such as source code, configuration files, sample data, YAML manifests, templates, or starter projects. By keeping all user-facing files under exercises/, the workshop environment stays organized and free of clutter from the home directory.
When this directory exists in the workshop files imported into the workshop session container, Educates treats it specially:
- Terminal working directory: Embedded terminals in the workshop dashboard start with
~/exercises as their current working directory instead of the home directory.
- Editor root directory: The VS Code editor opens on
~/exercises rather than the home directory, so users only see workshop-relevant files and are not distracted by hidden dot files or other home directory contents.
This special behavior is only triggered if the exercises/ directory already exists when the workshop session starts. It cannot be created later by workshop instructions — the directory must be part of the published workshop files.
Important: the directory must contain at least one file. Empty directories are not preserved when publishing workshop files to a Git repository or OCI image artefact. To ensure the directory is included, add a README.md with a brief note such as "Exercise files for this workshop" or a similar placeholder. Avoid using a .gitkeep file unless the workshop source is managed in a Git repository where that convention makes sense.
Because the exercises/ directory is always recommended, workshop instructions should never need to create it. File paths used in clickable actions for files under this directory must use the ~/exercises prefix (e.g., ~/exercises/deployment.yaml). The examples in the clickable actions reference files already follow this convention.
5. Generate workshop.yaml
Refer to references/workshop-yaml-reference.md for the complete workshop.yaml structure and options.
Generate the resources/workshop.yaml file based on the gathered details.
CRITICAL: Use the correct publish and workshop.files format.
The workshop.yaml MUST include these sections with this exact structure (substituting the actual workshop name):
spec:
publish:
image: "$(image_repository)/{workshop-name}-files:$(workshop_version)"
files:
- directory:
path: .
includePaths:
- /workshop/**
- /exercises/**
- /README.md
workshop:
files:
- image:
url: "$(image_repository)/{workshop-name}-files:$(workshop_version)"
IMPORTANT: Do NOT use spec.content.files — this is a deprecated format. Always use spec.publish and spec.workshop.files as shown above. The $(image_repository) and $(workshop_version) variables must be used exactly as shown to support local workshop publishing and deployment workflows. The spec.publish.files section controls which files are packaged into the published OCI image — only the listed paths are included, keeping the image small. The spec.workshop.files section specifies where to pull the files from at runtime; since the published image is already filtered, no includePaths is needed there. The spec.workshop.files array also supports Git and HTTP sources and can contain multiple entries overlaid in order — see the "Alternative File Sources" section in the workshop YAML reference for details.
Additional configuration:
- Set
metadata.name to the workshop name
- Set
spec.title and spec.description from gathered details
- Set
spec.duration to estimated completion time (e.g., 15m, 30m, 1h)
- Set
spec.difficulty to one of: beginner, intermediate, advanced, extreme
- Always include terminal with
enabled: true and layout: split
- Enable only the additional session applications the workshop requires
- Set
spec.session.namespaces.security.token.enabled to false by default (it is enabled by default for historical reasons)
- Only set
spec.session.namespaces.security.token.enabled to true if the workshop needs kubectl or uses the Kubernetes console
- Omit any applications that are not needed (do not include with
enabled: false)
6. Add GitHub Actions Workflow for Publishing (Only When Requested)
Skip this step during initial workshop creation unless the user explicitly asks to set up publishing of the workshop to GitHub container registry using a GitHub action. This step also applies when a user returns to an existing workshop and asks to add publishing support after the workshop has already been created.
This step applies only to standalone workshops that live in their own Git repository. If the workshop is part of a course containing multiple workshops, do not use this GitHub action — publishing for courses is handled differently.
When requested, create a .github/workflows/publish-workshop.yaml file in the workshop repository. Refer to references/workshop-publishing-reference.md for the complete workflow configuration, action parameters, and how publishing relates to the spec.publish section in workshop.yaml.
7. Add Local Docker Compatibility (Only When Requested)
Skip this step unless the user explicitly asks for the workshop to work on local Docker as well as in Kubernetes. This step also applies when a user returns to an existing workshop and asks to add local Docker support retrospectively.
Not all workshops are compatible with local Docker deployment. Workshops that require Kubernetes access, a virtual cluster, a session image registry, or that use environment.objects, session.objects, or request.objects in the workshop definition cannot run on local Docker.
When local Docker support is requested for a compatible workshop, the main change is appending the ingress_port_suffix data variable to all session proxy URLs — in both the workshop definition and workshop instructions. This variable is an empty string when on standard ports (Kubernetes) and includes the port number when on a non-standard port (local Docker), so adding it has no effect on Kubernetes deployment.
Refer to references/local-docker-deployment-reference.md for the full list of restrictions, where to apply the port suffix, and how to retrofit local Docker support onto an existing workshop.
8. Create the AI Assistant Instructions File
Skip this step if any of the following are true:
- An AI assistant instructions file (e.g.,
CLAUDE.md, AGENTS.md) already exists in the workshop root directory
- An AI assistant instructions file already exists in a parent directory (indicating the workshop is part of a larger project, such as a course created with the educates-course-design skill)
Create an AI assistant instructions file in the project root so that future AI interactions automatically know the project context and which skills to use. For Claude Code, this file is CLAUDE.md; other AI coding agents use different conventions (e.g., AGENTS.md).
The instructions file should contain:
- A pointer to
README.md for the project overview and description
- Skill references — when to invoke each skill:
- The educates-workshop-authoring skill for creating or modifying the workshop definition, instruction pages, and exercise files
- The educates-course-design skill for course planning, if this workshop is later incorporated into a multi-workshop course
Keep this file focused on AI-specific instructions and project-specific overrides. Do not duplicate content that already exists in README.md — reference it instead.
9. Create Workshop Instructions
Workshop instructions are placed in the workshop/content/ directory as Markdown files rendered by Hugo.
Guided Instruction Through Clickable Actions
By default, workshop instructions should provide a guided experience where all code interaction — viewing, running, and modifying — is driven through clickable actions. Learners should not be asked to type commands into the terminal or write code into the editor by hand. Instead, every interaction should use the appropriate clickable action (terminal:execute, editor:open-file, editor:replace-matching-text, etc.). This keeps learners focused on the concepts rather than on mechanics. If the person requesting the workshop explicitly asks for a different experience, adjust accordingly.
Refer to references/workshop-design-principles.md for the complete design philosophy, including guidance on how learners should view, run, and modify code, and how to structure the exercises/ directory as a pre-populated workspace.
Clickable Actions in Instructions
Workshop instructions use clickable actions — special fenced code blocks that let users execute commands, edit files, and interact with the workshop environment by clicking. Refer to references/clickable-actions-reference.md for the complete list of action types and detailed syntax.
Critical YAML safety rule for terminal commands: When generating terminal:execute clickable actions, always use YAML block scalar syntax (command: |-) if the command contains any characters that are special in YAML (colon, hash, curly braces, square brackets, etc.) or if the command spans multiple lines. This prevents the YAML parser from misinterpreting the command. For example:
```terminal:execute
command: |-
docker run --rm -p 8080:80 nginx:latest
```
Also ensure that shell commands use correct quoting — variable expansions containing paths with spaces should be double-quoted, strings with special characters should be properly escaped, etc. See the "YAML Syntax Safety" section in the clickable actions reference for detailed guidance and examples.
Critical YAML safety rule for editor actions with indented text: When generating editor clickable actions (editor:select-matching-text, editor:replace-matching-text, etc.) where the match, replacement, or text content starts with leading spaces — i.e., every line is indented — you must use a YAML block scalar indent indicator (e.g., |2, |4) to preserve the whitespace. Without it, the YAML parser strips the leading spaces, causing matches to fail or replacements to be inserted with incorrect indentation. For example:
```editor:replace-matching-text
file: ~/exercises/deployment.yaml
match: |2
containers:
- name: app
image: myapp:v1
replacement: |2
containers:
- name: app
image: myapp:v2
```
An alternative that avoids needing an indent indicator: expand the match to include enclosing context that starts at column 1. For example, instead of matching only an indented function body, include the function definition line (e.g., def my_function():) so the first line of the block scalar has no leading spaces and a plain | suffices. This works well for top-level functions but may not help for class methods or other constructs that are themselves indented. See the editor file actions reference for full details on indent indicators.
Critical YAML safety rule for trailing blank lines in text properties: When an editor clickable action's text or replacement property ends with one or more blank lines, those blank lines can be lost at two stages: YAML's default "clip" chomping strips trailing newlines from block scalar values, and Hugo's Markdown processor strips trailing blank lines from code fences. To preserve trailing blank lines, you must do both of the following: use the + keep chomping indicator on the block scalar (|+ instead of |), and add eot: true as the final property in the YAML. For example:
```editor:insert-lines-before-selection
file: ~/exercises/app.py
text: |+
# --- Section boundary ---
eot: true
```
The |+ tells YAML to keep trailing newlines, and the eot: true property (which has no meaning to the clickable action) ensures the blank lines are not trailing content in the code fence where Hugo would strip them. See the "Trailing blank lines in text properties" section in the clickable actions reference for detailed guidance.
Tracking Terminal Working Directory
When the exercises/ directory exists, each terminal session starts with ~/exercises as its current working directory — not the home directory. As you write workshop instructions, you must track the current working directory of each terminal at every point in the instructions. Any cd command in a terminal:execute action changes the working directory for all subsequent commands in that terminal.
Getting this wrong leads to commands that reference incorrect file paths. Either:
- Track the working directory and use correct relative paths. For example, if the terminal is in
~/exercises and you need to access ~/exercises/deployment.yaml, use deployment.yaml. If a previous step ran cd ~/exercises/myapp, then deployment.yaml would need to be ../deployment.yaml or you must use the full path.
- Use absolute paths to avoid ambiguity. For example, always use
~/exercises/deployment.yaml regardless of the current directory.
When the workshop uses the split terminal layout (two terminals), track the working directory of each terminal independently — a cd in one terminal does not affect the other.
Dashboard Tab Visibility
The workshop dashboard shows only one tab at a time on the right-hand side of the screen (the left side displays the workshop instructions). The Terminal tab is visible by default when a session starts. Users switch between tabs by clicking on tab headers or through dashboard:open-dashboard clickable actions.
This matters when writing instructions because certain actions implicitly change which tab is visible. A terminal:execute action switches to the Terminal tab, hiding any other tab (such as a web application dashboard) the user was viewing. If the workshop uses a custom dashboard tab (e.g., for a web app accessed via the session proxy), you must explicitly guide the user back to that tab after running terminal commands so they can see the result.
Dashboard tabs are iframes with no URL bar — users cannot see or change the URL the tab is pointing at. Because the content is embedded in an iframe, the page title of the embedded web application is also not visible anywhere in the browser — the browser tab/window title is dictated by the topmost parent page, not the iframe content. To change what a dashboard tab displays, use dashboard:reload-dashboard with a new url. Use dashboard:open-dashboard (not dashboard:reload-dashboard without a URL) when you only need to make a tab visible without reloading its content. When a workshop will change a tab's URL during the instructions, prefer creating the tab dynamically with dashboard:create-dashboard rather than pre-defining it in spec.session.dashboards, so the initial URL is visible in context.
Refer to references/workshop-dashboard-reference.md for detailed patterns and examples for handling tab switching in workshop instructions.
Data Variables in Instructions
Workshop instructions should be parameterized using data variables rather than hardcoding session-specific values. Educates provides data variables for the session namespace, ingress domain, session hostname, and many other context-specific values. Use the Hugo param shortcode to insert them:
Deploy the application to the `{{< param session_namespace >}}` namespace.
Data variables also work inside clickable actions:
```terminal:execute
command: kubectl get pods -n {{< param session_namespace >}}
```
In terminal commands within clickable actions, you can alternatively use the equivalent uppercase environment variable (e.g., $SESSION_NAMESPACE) since the terminal shell has these set automatically. Refer to references/data-variables-reference.md for the complete list of available data variables and which contexts they can be used in.
Page Structure
Each page requires YAML frontmatter with at least a title property:
---
title: Page Title
---
This is the introductory paragraph for the page. It appears immediately
after the frontmatter with no heading.
## First Section
Content for the first section...
## Second Section
Content for the second section...
Content guidelines:
- Use standard Markdown for page content
- Do NOT use a level 1 heading (
#) — the title in frontmatter automatically generates the page header
- Begin immediately with an introductory paragraph after the frontmatter
- Use level 2 headings (
##) and below for any additional sections
- Use admonition shortcodes to highlight important information:
{{< note >}} for tips, {{< warning >}} for cautions, and {{< danger >}} for critical warnings. See references/hugo-shortcodes-reference.md for syntax and usage guidance.
- Focus on the workshop topic, not the platform. Workshop instructions should teach the subject matter, not how Educates works. When the workshop requires platform-specific configuration (e.g., setting up a session proxy for accessing a deployed service, configuring ingresses, or using data variables), present these as natural steps of the exercise without drawing attention to Educates internals. Do not say things like "we will learn how Educates is configured" or "this is how Educates handles ingress" — unless the workshop is specifically about using the Educates platform itself. The overview, summary, and learning objectives should describe what users will learn about the topic, not about the workshop infrastructure supporting it.
File Naming Convention
Use a numeric prefix for ordering pages: nn-page-name.md
Recommended structure:
00-workshop-overview.md - Introduction and objectives01-first-topic.md - First instructional page02-second-topic.md - Continue incrementing for core content99-workshop-summary.md - Wrap-up and next steps
Reserve 00- for the overview and 99- for the summary. Core instructional pages start at 01- and increment.
Page Bundles
For simple pages, use a single .md file. For pages with embedded images local to that page, use a page bundle instead:
workshop/content/
├── 00-workshop-overview.md
├── 01-first-topic/
│ ├── index.md
│ └── diagram.png
├── 02-second-topic.md
└── 99-workshop-summary.md
A page bundle is a directory containing index.md plus any associated assets (images, etc.). For detailed guidance on including images in workshop pages, see references/images-in-workshop-pages.md.
10. Verify Workshop Definition
After generating resources/workshop.yaml, verify the following critical items:
Required sections exist:
Deprecated formats NOT used:
Application settings:
11. Verify Workshop Instructions
After generating workshop instruction pages, verify the following:
Terminal working directory correctness:
File path consistency:
Dashboard tab visibility:
YAML syntax in editor actions:
Guided instruction:
Content focus:
12. Update Planning Documents (Course Workshops Only)
When the workshop is part of a course that has planning documents (i.e., a planning/ directory exists containing workshop plans), check whether any significant changes made during implementation need to be reflected back in the planning documents. The goal is to keep the workshop plan accurate so that planning for subsequent workshops starts from correct information.
What to update:
- Page changes: If you changed a page's topic or title, rename the page file to match the new content and update the workshop plan's page listing accordingly.
- Approach changes: If you changed the overall approach — switched technique, simplified or restructured exercises, chose a different library or tool — update the workshop plan to describe what was actually built rather than what was originally planned.
- Renamed files: If you renamed a page file, check that no other workshop pages or planning documents still reference the old filename.
- Sequential workshop impact: For sequential workshops, check whether the next workshop's plan or implementation references anything that changed — its "Connection to Previous Workshop" section, exercises, or narrative hooks may assume the original approach.
Use the educates-course-design skill to make these updates — it owns the planning document format and ensures changes follow the correct structure. The important thing is that the workshop plan ends up reflecting the as-built design rather than the original draft.
Reference Guides
For detailed guidance on specific topics, see:
- Workshop Design Principles - Guided experience philosophy, the no-manual-typing rule, and guidance on how learners should view, run, and modify code
- Workshop YAML Reference - Complete workshop.yaml structure and options
- Images in Workshop Pages - How to include images using page bundles and static assets
- Clickable Actions Reference - Index of all clickable action types, YAML syntax safety guidance, and links to category-specific references in references/clickable-actions/
- Workshop Tools Reference - Command-line tools available in the workshop environment, including utilities for JSON/YAML processing, Kubernetes management, container handling, and load testing
- Kubernetes Access Reference - Namespace isolation, session namespace references, and pod security policies for workshops with Kubernetes access
- Data Variables Reference - Complete list of data variables for parameterizing workshop instructions, terminal commands, and workshop definitions
- Workshop Dashboard Reference - Dashboard layout, single-tab visibility behavior, and patterns for guiding users between tabs in workshop instructions
- Workshop Image Reference - Container image selection for workshops, including pre-built JDK and Conda images
- Java Language Reference - JDK image selection, Maven/Gradle build commands, project layout, and Spring Boot patterns for Java workshops
- Python Language Reference - Python version management, uv/pip package installation, project layout, and web framework patterns for Python workshops
- Workshop Setup Reference - Setup scripts, environment variables, background services, and terminal customization for the workshop container
- Hugo Shortcodes Reference - Admonition callouts (note, warning, danger), pathway conditional rendering, and custom shortcodes for workshop instructions
- Session Objects Reference - Pre-creating Kubernetes resources per session, shared environment objects, request objects, and workshop container resource configuration
- Workshop Publishing Reference - How to add a GitHub Actions workflow for publishing a standalone workshop to GitHub container registry using the Educates publish-workshop GitHub Action. Consult this when a user asks to set up publishing for a workshop, whether during initial creation or when adding it to an existing workshop later.
- Local Docker Deployment Reference - Restrictions and required changes for workshops that need to run on local Docker as well as in Kubernetes. Consult this when a user asks for local Docker compatibility, whether during initial creation or when retrofitting it onto an existing workshop.
- Local Cluster Deployment Reference - How to publish, deploy, update, and delete workshops on a local Educates cluster created with
educates create-cluster. Consult this when a user asks to publish or deploy a workshop to a local cluster. Do not run these commands automatically — only when the user explicitly asks to manage a workshop in the local cluster.
Skill Version
When asked about the skill version, read the VERSION.txt file and report its contents to the user.
Getting Help
For more information, visit the Educates documentation: https://docs.educates.dev/
