A toolkit of Claude Code skills for doing economic theory with LLMs: math-proof (single-pass proving), codex-math (adversarial verification), and co-math (multi-agent proof projects). With a VCG-for-grade-inflation case study.
Agent
Specialized sub-agent for the AI co-mathematician system. Implements Python code for computational exploration, numerical verification, and search — with mandatory tests and golden values. Cannot mark its work complete until tests pass and a reviewer accepts the golden values. Use when the project-coordinator…
Agent
Specialized sub-agent for the AI co-mathematician system. Attempts to formalise a stated lemma in Lean 4 and verify it with lake build. A successful Lean build is the strongest form of "proven" the system supports — accepted by paper-reviewer with only plumbing checks, not mathematical re-checking. Use when the…
Agent
Specialized sub-agent for the AI co-mathematician system. Performs literature searches and produces verified, cited workstream reports for a math research project. Use when the project-coordinator dispatches a literature workstream, when a workstream needs to confirm prior art, or when a claim in paper.tex needs…
Agent
Adversarial reviewer agent for the AI co-mathematician system. Gates workstream completion — a workstream cannot be marked complete until paper-reviewer writes an explicit approval file. Cross-checks references resolve, that proofs have no leftover \unproven blocks (or are properly flagged), that code outputs match…
Agent
The front door of an AI co-mathematician research project. Reads goals.md, talks to the user as a sounding board, formalizes research intent, dispatches workstreams to specialized sub-agents (literature-reviewer, prover, coder), and uses progressive disclosure to filter low-level chatter from the user. Use when the…
Agent
Specialized sub-agent for the AI co-mathematician system. Drafts mathematical proofs into paper.tex with strict discipline — every step is justified, cited, or explicitly marked \unproven. Never hand-waves. Use when the project-coordinator dispatches a proof workstream, or when a specific lemma/theorem in the paper…