Skill Tutor

A Claude skill that transforms any other Claude skill — or any domain — into a personalized human learning path.

When you ask:

/skill-tutor how can I learn to do {the skill name you want to learn} on my own?

Skill-tutor diagnoses what you already know, identifies learning gaps, builds a learning path calibrated to your starting point, and tries to deliver pedagogical content link by link with spaced repetition, interleaved challenges, and conceptual bridges.

Installation

Claude.ai (web / desktop)

1. Download this repo as ZIP (Code → Download ZIP)
2. Go to Customize → Skills → "+" → Create skill
3. Upload the ZIP

Claude Code

gh skill install malberich/skill-tutor

Or manually:

git clone https://github.com/malberich/skill-tutor.git
cp -r skill-tutor ~/.claude/skills/skill-tutor

API

See the Skills API quickstart for integration via the Anthropic API.

What it does

1. Identifies the target — finds the relevant installed skill or domain
2. Maps the capabilities — extracts every atomic capability needed to replicate the result
3. Diagnoses the learner — uses evidence-based questions (not self-assessment) to find real gaps
4. Builds a learning path — presents a skeleton with time estimates and early wins
5. Delivers content — link by link, using a five-section structure with recalls, interleaved challenges, and conceptual bridges

Key design decisions

• Evidence-based diagnosis: asks "have you ever...?" not "do you know X?" — because people are unreliable judges of their own competence
• Conversational pacing: one content block at a time with specific continuation options
• Spaced repetition: recalls at increasing intervals (links 3, 5, and final summary)
• Interleaved challenges: optional cross-link questions that activate retrieval and transfer
• Conceptual bridges: explicit connections between capabilities that combine into larger frameworks
• Dynamic adjustment: if the learner shows more mastery than expected, the path adapts

Structure

skill-tutor/
├── SKILL.md                              # Entry point
├── LICENSE.txt                           # MIT
├── README.md
├── .gitignore
├── references/                           # Claude loads on demand
│   ├── delivery-protocol.md
│   ├── pedagogical-structure.md
│   ├── style-guide.md
│   └── bloom-guide.md
├── docs/                                 # Human-facing documentation
│   ├── testing-guide.md
│   └── eval-results.md
└── evals/                                # Test suite (not loaded by Claude)
    ├── evals.json                        # 42 tests, skill-creator format
    └── skill_tutor_test_dataset.xlsx     # Same tests as spreadsheet

Testing

See docs/testing-guide.md for the full evaluation framework: 42 test cases across 10 categories, priority order, and instructions for manual and simulated evaluation.

See the evals folder for the specific testing process being built with the skill-creator skill provided by Anthropic.

License

MIT — see LICENSE.txt.