THE ENGAGEMENT
What we reviewed
Bench is an opportunity markets protocol built on Solana using Anchor, with Arcium for confidential compute on the reveal/stake logic. We ran a preaudit on their program ahead of a full audit: AI-assisted scan plus human validation on the codebase at commit 2dc1caf4. We triaged every flagged candidate, cut false positives, and recalibrated severity based on real exploitability. Found 11 validated issues (0 critical, 0 high, 4 medium, 7 low), mostly reward and lifecycle edge cases in reveal/finalization, giving Bench a clear punch list to fix before committing to a full audit.
SCOPE
Areas covered
HIGHLIGHTS
What the review found
Preaudit run ahead of a full audit: AI-assisted scan with human validation of every candidate
False positives triaged out and severity recalibrated against real exploitability
Delivered a prioritised punch list before committing to a full audit
Findings by severity
Ship Safely.
Whether you're pre-launch or scaling post-TGE, we'll cover every layer; Code, infrastructure and/or operations.