qchem-leak-screen v0.1.0 flags physically impossible AI-predicted quantum properties on CPU

A new open-source tool called qchem-leak-screen v0.1.0 has been released to validate quantum chemical properties predicted by AI models, such as dipole moments, HOMO/LUMO energies, and band gaps, against hard physical laws. The tool works entirely on CPU without requiring a GPU or running any DFT calculations, operating solely on values already output by other models. It applies four physically derived rules covering atomic valency, Koopmans' theorem, point-group symmetry, and physical range checks, flagging violations with an explicit rule ID, severity level, and reason. Results can be exported in JSON, Markdown, or SARIF 2.1.0 format, making it directly compatible with GitHub Code Scanning pipelines. The project ships with 34 passing tests, 88.31% code coverage, and processes each molecule in under 50 milliseconds, with its sole runtime dependency being the typer library.