math-olympiad

Solves competition math problems through multi-agent verification and calibrated abstention.

Updated Jul 10, 2026
One-click install
npx skills add https://github.com/AvaTar-ArTs/.Agent-skills --skill math-olympiad-avatar-arts
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: math-olympiad
Source: https://github.com/AvaTar-ArTs/.Agent-skills/tree/main/skills/math-olympiad-solver
Command: npx skills add https://github.com/AvaTar-ArTs/.Agent-skills --skill math-olympiad-avatar-arts

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Competition math problems (IMO, Putnam, USAMO, AIME) demand rigorous proofs where subtle errors are common and a single wrong step invalidates the answer. This Skill applies a structured multi-stage workflow that catches flawed reasoning before presenting a solution. ## Core Features & Use Cases - Parallel Candidate Generation: Multiple solver agents attack the problem from different angles, each refining its approach through self-verification cycles. - Adversarial Verification: Independent verifier agents attack cleaned proofs using pattern-based checks for domain errors, invalid hypotheses, and missed edge cases, with asymmetric voting (4 holds to confirm, 2 holes to refute). - Revision and Calibrated Abstention: A reviser agent fixes flagged proofs up to three times; if verification still fails, the Skill honestly reports no confident solution instead of guessing. - Use Case: Given a USAMO proof problem, the Skill generates candidate proofs, adversarially verifies them, revises any holes, and delivers a polished final proof, optionally compiled to PDF with LaTeX. ## Quick Start Solve this competition math problem with rigorous verification and present a clean final proof.

Frequently Asked Questions about math-olympiad

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I solve IMO or Putnam problems with AI verification?

Submit the competition problem and the Skill launches multiple solver agents in parallel, then adversarially verifies each cleaned proof with independent verifiers. Only solutions passing asymmetric voting are presented as final.

How does adversarial proof verification work for math problems?

Verifier agents receive only the cleaned proof, without the solver's reasoning trace, and attack it using pattern-based checks for domain errors, invalid hypotheses, and unhandled edge cases. Four holds confirm a solution while two holes refute it.

Can it handle numeric-answer contests like AIME?

Yes, for numeric-answer problems such as AIME the Skill uses a simplified workflow focused on answer verification rather than the full proof-verification pipeline used for proof-based contests.

What happens when no correct solution is found?

The Skill practices calibrated abstention: after up to three revision cycles fail verification, it states no confident solution was found and explains what was attempted and where the proof breaks, rather than guessing.

What are the limitations of AI competition math solving?

The Skill performs no web searches or literature reviews and relies on internal reasoning plus bounded local computation like modular arithmetic. It abstains on problems requiring deep specialized domain knowledge or heavy computation.