TankTechnology
<!-- clrs-audit-2026-09-08-chapter-31 --> <!-- clrs-repair-2026-09-08-chapter-31 --> ## Resolution This MAJOR chapter-audit finding is resolved. The accepted repair is [dd0815b25dc0](https://github.com/TankTechnology/CLRS-Lean/commit/dd0815b25dc06325fbd12f35862367ef44797596) and was integrated by [PR #377](https://github.com/TankTechnology/CLRS-Lean/pull/377). The repair completes this acceptance objective: **Connect Miller-Rabin counts to executed tests and repeated-trial probability.** The [chapter-repair index](https://github.com/TankTechnology/CLRS-Lean/blob/main/docs/plans/2026-09-08-chapter-repairs/index.md) records the implementation boundary and focused evidence. The [unified verification record](https://github.com/TankTechnology/CLRS-Lean/blob/main/docs/plans/2026-09-08-chapter-repairs/verification.md) records the 10,768-job library build, all 35 chapter trust checks, focused regressions, repository checks, and Pages-equivalent local validation. ## Audit basis and claim boundary The finding came from the September 8 fourth-edition source review at baseline `c8b074e961fb204ab50b9ea123b864260dafde38`. Source definitions, theorem premises, and kernel-checked counterexamples were inspected directly. Textbook equivalence remains `NOT-INDEPENDENTLY-VERIFIED` because the book was not checked independently page by page. Closing this issue means its stated code, proof, documentation, and model-boundary criteria are satisfied. It does not extend charged-operation results into RAM or bit complexity, or claim that every textbook theorem and exercise is formalized.