Skills
Skills are written procedures your assistant reads at the right moment: how to onboard a teacher, how to
drive a lesson without answering for the learner, how to build a self-study lesson. They come with the
repository in skills/. A skill brings no model, account or server of its own: connect the Studium server
first (Harnesses).
You do not need a skill to learn a topic in a conversation. When the server connects, it gives the
assistant its instructions, and the learn and review prompts start the two everyday flows.
Available skills
| Skill | Use it to | Example request |
|---|---|---|
studium-student |
install or connect Studium, set up Copilot Student or another assistant, start a first lesson | "I have Copilot Student; help me start the course." |
studium-teacher |
install, connect an assistant, draft a lesson from literature, review it before a pilot | "Help me prepare my first course from this article." |
studium-harness |
run the lesson loop over MCP, HTTP or the file channel: present scenes, collect the learner's answers, run capsules honestly | "Drive this session and wait for my answers." |
studium-self-learner |
turn references you chose (a web page, a PDF, a text) into a self-study lesson with the Python packet tools, then play it | "Build a lesson from this chapter, then play it with me." |
studium-from-conversation |
turn the conversation in progress, or a past session, into a short lesson that quotes it, with a concept map | "Make a short lesson from this conversation so I can check I understood why the gate sent me back." |
studium-concept-map |
show where you stand on each concept of a lesson, and export the map as a page to explore, with what each concept rests on | "Where am I in this lesson? Then export the map as a page." |
studium-suggest |
let the assistant offer the right Studium feature at the right moment, once, and wait for your yes | "Tell me when a short lesson or a review would help, and ask first." |
studium-activity-templates |
choose and adapt an interactive activity tied to a scene | "Let me explore this result with a template." |
studium-visual-support |
inspect source photographs or request an adapted image, map, audio or video through your assistant | "Prepare a visual aid with a text equivalent." |
studium-learner-report |
export a recorded activity trace or experiment journal as a local report | "Prepare my report without my private texts." |
studium-observatory |
run and read the Observatory, the read-only local view of what Studium produced | "Show me the decisions and evidence behind this course." |
studium-contribute |
report a difficulty with Studium itself or ask for a missing feature: the assistant drafts the text, you post it | "Help me report what just happened; I will post it myself." |
Use them in a checkout
The repository links the skills into .claude/skills/ (Claude Code) and .agents/skills/ (Codex and other
clients that read that folder). Clients that recognise these folders can load a skill on their own;
discovery depends on the client, its version, whether you trusted the project, and whether your checkout
kept the symlinks. If your client does not list a skill, studium-observatory for example, name its file.
In Codex, call a skill by name:
Use $studium-student to connect my assistant and start the lesson my teacher gave me. Wait for my own answers.
In any client, the portable way is explicit:
Read skills/studium-teacher/SKILL.md in this repository and apply it to my teacher onboarding.
Replace the name with the skill you need. The skill resolves its real folder to find the guides and tools it refers to.
Use them from another folder
To use the skills outside the checkout, link each skill folder into the skills folder your client reads,
for example ~/.claude/skills for Claude Code or ~/.agents/skills for Codex:
ln -s /path/to/studium/skills/studium-student ~/.claude/skills/studium-student
Point the links at a durable checkout, not a temporary worktree, and keep any entries already there. Do not
copy SKILL.md alone: it refers to guides and tools in the repository. After an update, open a new session
if your client caches its skill list. On Windows without symlinks, name the source file in your prompt.
What a skill does not prove
A loaded skill does not prove that the Studium server is connected: check with get_server_info
(Harnesses). An automated demo does not prove that anyone learned.
A request to onboard does not authorise a purchase, sharing with students or publishing data; the skills
ask before such steps and reuse the permissions you already gave for local ones.
Related pages: Report a problem or contribute, Activity templates, Media and visual support, Learner report, Observatory.