What problem does it solve? Finding new, undiscovered bugs in the Cranelift compiler is difficult because miscompilations and logic errors can hide in optimization passes, lowering rules, and backend-specific ABI code, yet they can lead to sandbox escapes in Wasmtime. This Skill guides a systematic audit of the Cranelift codebase to surface real, demonstrable bugs. ## Core Features & Use Cases - Structured Bug Hunting: Defines what counts as a bug (incorrect optimizations, incorrect lowering rules, backend ABI bugs, logic errors) and excludes known or already-fixed issues. - Reproducible Bug Reports: Requires every finding to include a *.clif test case executed with clif-util, including differential execution against the Cranelift interpreter and cross-architecture testing via QEMU userspace emulation. - Organized Audit Output: Produces numbered report folders (reports/NNN-description) with human-readable report.md files plus intermediate findings in reports/findings/, preserving results from previous audits. - Use Case: A security researcher auditing Wasmtime before a release uses this Skill to methodically review Cranelift optimization passes, confirm a miscompilation with a clif test on x86 and aarch64 via QEMU, and file a complete bug report. ## Quick Start Audit the Cranelift codebase for new miscompilation bugs and write reproducible clif test cases for any findings into the reports folder.
