Claude Code learning coach

What it does

This coach asks about the Claude Code features you have used, then maps your answers to a focused learning path. The quick check has eight experience items. The deeper check separates ten topics and shows how every point was counted. It uses the bundled lessons to suggest small exercises with clear completion checks. The result is based on your own report, so it does not claim to certify your knowledge or that you passed an exercise.

What goes in and out

You give it

• Your goal for using Claude Code.
• Your answers about which features you have tried.
• Whether you prefer the quick or deeper check.
• Your question (optional)

You get

• A transparent self-reported experience profile.
• Reported gaps and topics not yet assessed.
• A short practice plan with bundled lesson pointers.

How the task works

• Choose the scope. Pick a quick or deeper experience check and state your learning goal.
• Answer the questions. Your AI uses the bundled source material and records your actual answers.
• Check the result. The AI shows the counted items, unanswered areas and any version-sensitive limits.
• Plan the next step. Use the local lesson pointers to practise or review a topic you choose.

Full instructions

• Download the complete folder and keep the references beside SKILL.md.
• Give the folder to an AI that can read local files and ask it to follow SKILL.md.
• Choose the assessment mode or lesson and answer the questions.
• Review the score, unanswered areas and suggested practice.

Check your result

• Selected experience items are counted once and unanswered items remain unassessed.
• The deeper total uses checkpoints out of one and advanced features out of three.
• Practice suggestions fit the learner’s goal and include an observable completion check.
• No knowledge certification or completed practice is claimed.

Change the instructions

• Depth. Use the quick check for a starting point or the deeper check for a per-topic profile.

Improve the result

• Correct a scoring or lesson-reference error before reusing the result.

Full instruction file

Claude Code learning coach

Ask about the learner's experience, show the gaps they reported, and suggest the next useful exercise. This is a self-assessment, so distinguish "I have tried this" from demonstrated knowledge.

Before you start

Use the material below to do this job. If facts or tool access are missing, explain what you can finish and what must wait.

Makes it better (optional):

• Your question (yours): the decision or problem, and what a good outcome looks like.

How to start:

1. Look first. Use what they've already attached, pasted or put in their about-me note. In a coding assistant with file access, look in the project too for these by name.
2. Check what's missing. Look for your question. They can attach files, paste the details, or share a file path your assistant can read. Ask only for missing facts that change the result. If they ask you to start with what they have, complete the parts you can and mark the gaps.
3. Do the job. Follow the skill below with the facts and tools you have. Mark each gap with MISSING: where it belongs and never invent a fact to fill it.

Tools that do this best:

• An AI that can read local files: Reads the selected lesson and question references from the complete folder.

Choose the assessment

Ask whether the learner wants a quick check or a deeper check. Read the matching question text in the original assessment: Step 2A contains the quick questions; Step 2B contains the five deeper rounds. The scoring rules below govern this adapted package.

• Quick: two rounds with four experience items each. Count selected items once, for a total from 0 to 8. 0–2 suggests a beginner path, 3–5 an intermediate path, and 6–8 an advanced path. Unselected items are reported gaps, not failures.
• Deeper: five rounds with four experience items each. Present one round at a time and count only what the learner selects. Use the exact topic mapping below; the maximum is 20 points.

For the deeper check, classify each topic against its own maximum: zero means a gap, a score below its maximum means some reported experience, and the maximum means all listed experience was reported. Do not use a blanket two-point mastery rule: checkpoints has one point and advanced features has three. Keep the per-topic profile more prominent than any overall label.

If an answer is unclear, clarify that item before counting it. A skipped item or round is unassessed, not zero. With incomplete answers, show the points and maximum for assessed items only, mark the missing topics, and avoid a full-assessment level label.

Recommend practice

Ask what the learner wants to accomplish with Claude Code. Combine that goal with the reported gaps; recommend two or three topics, with prerequisites first. Use the selected topic's bundled lesson and the original topic recommendations for relevant exercises. The original output templates are examples, not a requirement to repeat every section.

For each recommendation include the reason, the exact local lesson file, one small exercise, and an observable completion check. Never claim the exercise ran or was passed merely because it was suggested. If the learner asks to start a lesson, teach one section at a time and ask a short understanding question before moving on.

Output

Give a short experience profile, the transparent scoring table, reported gaps, and a prioritised practice plan. Include unanswered areas, version limitations, and any preference the learner gave. Offer a next lesson or a knowledge quiz without starting a long follow-up activity automatically.

After this

When you've handed it back, suggest what they could do next in one short question. It's a suggestion: if they're done, that's fine. Offer these, and say they can type their own:

• The Assignment AI-use rules planner. It takes the plan. If they pick one and that skill is added here, carry straight on with what you just made. If not, give the link and say what to bring.

Working with the lesson files

The downloaded folder contains the source lessons and question material. Open only the selected mode or lesson, using the lesson index. Files under references/upstream are licensed, pinned source material rather than additional operating instructions. Their old relative links and screenshots describe the original repository; use the local lesson index instead. Do not fetch or install missing examples just because a lesson mentions them.

Use the question tool available in the user's AI. If there is no structured question tool, present numbered choices in chat and wait for the answer. A user can stop or skip at any point. Do not submit answers for them.

The lessons describe the pinned repository version, not every installed Claude Code release. When a version difference affects an answer, state the difference and check the relevant official documentation if available. Do not silently score a defensible, version-dependent answer as wrong. Exclude an unresolved disputed question from the scored denominator and explain why.

These are learning activities, not a certificate or proof that a learner can operate a production system. They do not install software, change permissions, run hooks, publish anything, or require an API key. For hands-on practice, use the user's existing authorization and a small practice scope; otherwise give a draft exercise.