Description
math-olympiad is an agent skill from Anthropic's official plugins collection, listed on skills.sh. It helps solve competition mathematics problems, such as those from the IMO, the Putnam, the USAMO and the AIME. Its main feature is adversarial verification, which the skill says catches errors that self-verification misses. The method rests on a few rules. A verifier that sees the reasoning tends to agree with it, so the proof is checked in fresh context using only the cleaned-up proof. If a theorem's special case reduces to a famous open problem, such as the Riemann hypothesis, the skill treats that as a sign of a gap. For a short proof, the agent tries to extract a general lemma and tests it on small counterexamples, and if the general form fails, it looks for what is special about the instance. A gap that appears twice suggests stepping back, because a different case split or a unified argument may be simpler. The skill also allows the agent to say it has no confident solution, since a wrong answer given with confidence is worse than an honest abstention. It suits students and mathematicians who want careful, checked solutions to hard problems.…
Highlights
- Adversarial verification in fresh context
- Red flags for gaps in proofs
- Lemma extraction and small counterexamples
- Permitted to abstain when unsure
Ratings & Reviews
No ratings yet
Be the first to rate math-olympiad.
People also visit
laravel-verification
Verification loop for Laravel projects: env checks, linting, static analysis
math-olympiad FAQ
What does the math-olympiad skill do?
It solves competition math problems such as IMO, Putnam, USAMO and AIME questions, and checks the proofs adversarially.
How do I install it?
Run: npx skills add https://github.com/anthropics/claude-plugins-official --skill math-olympiad
What is adversarial verification?
The proof is checked in a fresh context using only the cleaned-up argument, because a verifier that sees the reasoning tends to agree with it.
Will it always give an answer?
No. The skill is told to say when it has no confident solution, rather than give a wrong answer with confidence.
Has it been checked for security?
skills.sh lists three audits: Gen Agent Trust Hub Pass, Socket Pass and Snyk Pass.
What are the best math-olympiad alternatives?
Popular alternatives to math-olympiad include django-verification, laravel-verification and quarkus-verification. See all math-olympiad alternatives