math-proof

Two skills for hard mathematics problems, each ending in a self-contained proof.md: /math-proof:solo has the session work the problem itself in stages with a notes file; /math-proof:siege works on it for hours in rounds of judge and worker sub-agents and says plainly what is and is not proved.

Install

If the official marketplace isn’t added yet, run

Category: Math
Author: Anthropic
GitHub: anthropics/claude-plugins-official

Good for

Work on hard, research-level math problems, ending in a self-contained proof.md
Use /math-proof:solo to work the problem in your session in stages with a notes file
Use /math-proof:siege to run judge and worker sub-agents in rounds for hours
Get a Status section that says plainly what is and is not proved

Getting started

Run /plugin install math-proof@claude-plugins-official in Claude Code
Try the lighter /math-proof:solo first, with the problem stated in full
Move to /math-proof:siege if needed, checking /usage before you start
Read proof.md critically; results cited from the literature are quoted from memory

Before you install

siege runs dozens of sub-agents, over a hundred if all fourteen rounds run; on an API key you can cap spend with --max-budget-usd in the unattended form. Web tools are switched off during a run.

Summarized by AI from the project docs; check the official docs before relying on it.

In the news

10/23, 11:58Research
Anthropic develops AI math proof for bug-free code
Anthropic is developing a method that uses AI to perform formal verification, mathematically proving that software has no bugs. The approach aims to achieve 'bug-free software' in response to errors caused by high-volume AI-generated code.
GIGAZINE

Similar tools

math-olympiadOfficial plugin · Math
Solve competition math (IMO, Putnam, USAMO) with adversarial verification that catches what self-verification misses. Fresh-context verifiers attack proofs with specific failure patterns. Calibrated abstention over bluffing.