An adaptive math tutoring skill for Claude. It generates practice problems on any math topic, evaluates student answers, and provides progressive hints and walkthroughs calibrated to the student's level. A persistent student profile tracks strengths, weaknesses, and common mistakes so difficulty adjusts over time.
- Any math topic — from basic arithmetic to transfinite ordinals. For advanced or specialized topics, the tutor searches the web for reference material to ensure problems are genuinely level-appropriate.
- Adaptive difficulty — starts at a reasonable level and adjusts based on student performance. Batch sizes flex too: more problems for straightforward topics, fewer for complex ones.
- Progressive hint system — three escalating levels when a student is stuck: a conceptual nudge, a concrete first step, and a full walkthrough.
- Answer verification — the tutor solves each problem independently and verifies answers using multiple approaches before evaluating the student's response.
- Persistent student profile — tracks topics covered, strengths, areas for improvement, common mistake patterns, and session history in a markdown file that carries across sessions.
- LaTeX notation — all math expressions use LaTeX formatting for clean, readable presentation.
- Post-batch assessment — after each set of problems, the tutor summarizes performance and recommends next steps: more practice, higher difficulty, a new concept, or revisiting prerequisites.
Install the .skill file from the dist/ directory in Claude Code or Cowork.
math-tutor/
├── SKILL.md # Main skill instructions
├── references/
│ └── student-profile-template.md # Template for new student profiles
├── dist/
│ └── math-tutor.skill # Packaged skill file
└── evals/
└── evals.json # Test cases and assertions
When a student asks to practice a math topic, the tutor:
- Reads (or creates) a student profile to understand their level
- Asks whether they have a worksheet or want problems generated from scratch
- Presents problems one at a time with LaTeX formatting
- Evaluates each answer — celebrating correct ones and offering structured help for incorrect ones
- After each batch, gives an assessment and suggests what to work on next
- Updates the student profile with observations from the session
MIT — see LICENSE.