math-olympiad — Claude Code plugin: olympiad math solver
Solves IMO, Putnam and USAMO problems and checks proofs with separate agents that never see the reasoning trace.
/plugin install math-olympiad@claude-plugins-official
/plugin marketplace add anthropics/claude-plugins-officialWhat math-olympiad does
The plugin targets the weak spot of self-verification: a checker that sees the reasoning tends to agree with it. Here the verifier gets only the clean proof and attacks it with specific failure patterns — not "is this correct?" but prompts like "does this accidentally prove RH?" or "extract the general lemma, find a 2x2 counterexample".
When no confident solution exists, it says so instead of bluffing. Once verification passes, a presentation pass produces clean LaTeX or PDF output.
The skill auto-triggers on "IMO", "Putnam", "olympiad" and "verify this proof". It installs from Anthropic's official marketplace and ships under Apache-2.0.
What it helps with
- check an olympiad proof for hidden gaps
- get clean LaTeX output after verification passes
- tell a real solution apart from a plausible bluff
- work through an IMO or Putnam problem step by step
- find a counterexample to an over-general lemma
Example prompts for the agent
- Solve this IMO problem: [statement]
- Verify this proof: [text]
- This is a Putnam problem, format the solution in LaTeX
How to install math-olympiad
Plugin marketplace
/plugin install math-olympiad@claude-plugins-official/plugin marketplace add anthropics/claude-plugins-officialStep by step, with commands for each platform: Claude Code plugins: marketplaces, installing, and building your own
Questions about math-olympiad
How do I install math-olympiad in Claude?
Add the marketplace first: /plugin marketplace add anthropics/claude-plugins-official, then install the plugin: /plugin install math-olympiad@claude-plugins-official.
What does math-olympiad do?
It solves competition math problems and verifies proofs with context-isolated agents that hunt for specific failure patterns. If no confident solution exists, it abstains rather than bluffing.
How is its verification different from self-verification?
The verifier sees only the clean proof, never the reasoning trace, so it is not biased toward agreement. Checks run against named failure patterns instead of a generic correctness question.
Source text
math-olympiad · Apache-2.0
# math-olympiad
Competition math solver with adversarial verification.
## The problem
Self-verification gets fooled. A verifier that sees the reasoning is biased
toward agreement. arXiv:2503.21934 ("Proof or Bluff") showed 85.7% self-verified
IMO success drops to <5% under human grading.
## The approach
- **Context-isolated verification**: verifier sees only the clean proof, never
the reasoning trace
- **Pattern-armed adversarial checks**: not "is this correct?" but "does this
accidentally prove RH?" / "extract the general lemma, find a 2×2
counterexample"
- **Calibrated abstention**: says "no confident solution" rather than bluff
- **Presentation pass**: produces clean LaTeX/PDF after verification passes
## Validation
17/18 IMO+Putnam 2025 problems solved, 0 false positives, 2 novel proofs found.
See the skill's eval data in the
[anthropic monorepo](https://github.com/anthropics/anthropic/tree/staging/sandbox/sandbox/ralph/math_skills/eval_harness).
## Install
```
/plugin install math-olympiad@claude-plugins-official
```
## Use
```
> Solve this IMO problem: [statement]
```
The skill auto-triggers on "IMO", "Putnam", "olympiad", "verify this proof",
etc.
© Anthropic · Apache-2.0 · anthropics/claude-plugins-official
