Evidence at a glance
Based on the linked sources. The author's reported result and independent confirmation are separate checks.
- AI involvement
- Reported models: Claude Sonnet 5.5. Tools: Decompiler. The writeup below describes the work attributed to AI.
- Reported result
- Works. This is the directory's source-based outcome classification.
- Independent confirmation
- Not reported.
- Human work and limitations
- The target is a smart contract on the Base blockchain. The researcher confirmed using Claude for the decompilation process in response to a direct inquiry.
Original sources
What was done
A researcher identified a vulnerability in a closed-source Slice protocol vault that allowed unauthorized fund withdrawal. Claude Sonnet 5.5 was used to assist in the decompilation and analysis of the bytecode, pinpointing the flaw in under two minutes.
The researcher 0x9527 identified a vulnerability in a closed-source Slice protocol vault that resulted in a loss of approximately $41,000. The protocol suffered from a logic error where it trusted the return value of an external, caller-supplied contract to update internal accounting, enabling unauthorized vault drainage. Because the contract was not open source, the researcher performed reverse engineering by working through decompiled bytecode. The researcher reported using the free tier of Claude Sonnet 5.5 to assist in the decompilation and analysis process. According to the researcher, the model successfully pinpointed the vulnerable function within two minutes. While the bug itself was described as a textbook, embarrassingly basic flaw, the reverse-engineering effort was noted as the primary challenge in the investigation. See the original report at https://x.com/coffiasse/status/2112806669654925542.
