An AI agent can solve the hardest math competition problems. It uses a special two-step method. First it tries to solve the problem. Then a fresh verifier checks the solution for mistakes. This catches errors that normal self-checking misses.
The agent gives calibrated confidence in its answers. If it is not sure, it says no confident solution. It never makes up a false answer. The verification attacks the proof with specific failure patterns, not just general logic checks.
This skill works for problems from IMO, Putnam, USAMO, and AIME. It can also check if a proof is correct. It produces a clean PDF with LaTeX when verification passes.
Global
mkdir -p ~/.claude/skills/math-olympiadProject
mkdir -p .claude/skills/math-olympiadSource Repository
Opensource Guide Coachxixu-me/skills
Get practical open source coaching for starting, growing, and sustaining your project
Scaffold Exercisesmattpocock/skills
Easily scaffold course exercise directories with sections, problems, solutions, and automatic linting
Recipe Create Classroom Coursegoogleworkspace/cli
Create a Google Classroom course and invite your students easily
Dbs Learningdontbesilent2025/dbskill
Adaptive learning articles that evolve with your feedback to master any topic
Add Educational Commentsgithub/awesome-copilot
Turn any code file into a guided learning experience with clear comments
Vscode Ext Commandsgithub/awesome-copilot
Learn the guidelines for contributing commands to VS Code extensions
Story Long Analyzeworldwonderer/oh-story-claudecode
Deep analysis of hit web novels to help you write your own bestseller
Mentoring Juniorsgithub/awesome-copilot
Socratic mentoring that helps juniors find answers through guided questions