What problem does it solve? Drafting proofs for ML/AI theory often produces polished but flawed arguments with hidden gaps, unstated assumptions, or misapplied theorems. This Skill produces mathematically honest proof packages that either prove the claim, prove a corrected version, or explain exactly why the claim is not justified. ## Core Features & Use Cases - Feasibility Triage: Classifies every claim as PROVABLE AS STATED, PROVABLE AFTER WEAKENING / EXTRA ASSUMPTION, or NOT CURRENTLY JUSTIFIED before writing any proof. - Structured Proof Packages: Writes a PROOF_PACKAGE.md containing the exact claim, assumptions, notation, proof strategy, dependency map, numbered steps, and open risks. - Honest Blockage Reports: When a proof cannot be completed, it documents the exact blocker, missing lemma, or counterexample instead of fabricating steps. - Use Case: You are drafting a convergence lemma for an optimization paper and have a rough proof sketch. The Skill normalizes the claim, checks whether the sketch actually supports it, and writes a complete appendix-ready proof or flags the missing assumption. ## Quick Start Ask the assistant to prove your stated theorem or lemma under the listed assumptions and write the result into a proof package file.
