build-matching-decompilations
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill instructions in
SKILL.mdand thereferences/directory promote high-security standards for the domain, including SHA-256 pinning of toolchains, 'fail-closed' verification rules, and negative control testing to ensure verifiers work as expected. - [SAFE]: The
evals/directory contains deliberately flawed scripts (e.g.,verify.pyandprogress.py) that implement insecure practices like byte-masking, size-truncation, and incorrect progress counting. These are explicitly provided as training and evaluation targets for the agent to detect and fix, aligning with the skill's educational purpose. - [SAFE]: No malicious obfuscation, credential harvesting, or unauthorized network operations were detected. All external references point to reputable open-source tools within the decompilation community (e.g.,
objdiff). - [SAFE]: The skill includes instructions to strictly maintain original program behavior, including bugs and undefined behavior, which prevents the agent from introducing 'fixes' that would diverge from the reference binary.
Audit Metadata