How to use the Learn Everything Course Builder agent skill

What an agent skill is and what this one changes

An agent skill is a folder of instructions and optional supporting files that an AI agent loads when a request matches the skill's description. You do not call it with a special command; you ask for something and the agent decides whether the skill applies.

Learn Everything is built for learning. Once it is active, asking the agent to teach you a document or subject makes it first build a course map, then teach one chapter at a time instead of dumping everything at once. Each lesson alternates explanation, worked examples, quick checks, exercises and spaced review. Answers stay hidden until you attempt the exercise, unless you explicitly ask for them.

• Document mode: works from an uploaded PDF, paper, book chapter, slides or notes, and keeps page or section references.
• Topic mode: builds a curriculum from a subject you name, with research when the agent has web tools.
• Resume mode: picks up a previous course from a saved state file or a pasted state block.

Install the Learn Everything skill

There are two routes. Choose the one that matches how you use the agent.

For Claude Code or a compatible local agent, download the skill ZIP from this page and unzip it into your user skills folder, ~/.claude/skills/, or into a project's .claude/skills/ folder if you want it only for that project. For the Claude apps, upload the ZIP under Settings → Capabilities → Skills.

The skill's files are SKILL.md, references/session-formats.md, references/practice-and-mastery.md and agents/openai.yaml. Keep SKILL.md at the top of the skill folder so the agent can find it.

• Local agent: unzip into ~/.claude/skills/ (all projects) or .claude/skills/ (one project).
• Claude apps: Settings → Capabilities → Skills, then upload the ZIP.
• Check that the folder contains SKILL.md after unzipping.

Trigger the skill with natural requests

You do not need special syntax. The skill's description lists the kinds of requests it responds to, so describe what you want to learn in plain language. You can write in English or Chinese; the skill teaches in your language unless you ask for another.

Attach your material when you ask, or name the topic. The skill is designed for requests to learn, study, understand, master, review or practise a subject chapter by chapter, and for requests to continue earlier work.

• 'Teach me this' with a PDF or chapter attached.
• 'Learn this PDF' or 'help me study this paper chapter by chapter'.
• 'Build a curriculum from this URL' or 'teach me statistics from scratch' (topic mode).
• 'Continue my course' or 'pick up where we left off', with your saved state file or state block.
• Chinese triggers listed in the skill include '分章节教学', '带我学' and '接着上次学'.

Work through a session

Expect the agent to first ask one short diagnostic question, or offer a three-question quiz you can skip, if it does not know your level. It then gives you the target capability, a course map of 3 to 12 chapters, an estimated level, and the rule that lessons go one at a time.

Each lesson moves through orientation, objectives, explanation, a worked example, a misconception check, a brief recall question, practice and feedback. The agent decides whether to advance, review or make the next exercise harder based on your answer. For longer work, you can ask it to save progress so you can return later.

• Attempt the exercise before asking for the answer; the agent will not reveal it first unless you ask.
• In document mode, expect page or section references when it teaches source-specific content.
• Ask for a worksheet or exam if you want a larger set of questions; the default is 3 to 5 exercises per lesson.

Save progress and resume later

If the agent has file tools, it keeps a course state file at learn/<slug>/course-state.md. The file holds the chapter list with completion, a concept mastery table, a review queue and an error log. It is updated at chapter boundaries, not every turn.

If the agent has no file access, it gives you a compact state block at each chapter checkpoint. Copy it somewhere safe and paste it back in a new session. When you resume, the agent runs two to four warm-up questions, weighted toward concepts you found hard, then continues with the current chapter. If the source document is needed but you do not paste it again, the agent will ask for it.

• With file tools: look for learn/<slug>/course-state.md.
• Without file tools: save the state block the agent prints and paste it back to resume.
• Re-supply the original document when asked, so the agent can teach from the real text rather than summaries.