AI review evidence for scope, security, dependencies, CI, and payout readiness.
The AI gate reports evidence and manual review areas. It never claims absolute safety or final authority; it helps sponsors decide whether a bounty can move toward acceptance.
Competing claim reports scored by the specialized audit agent.
PR gates waiting for claim evidence, AI review, or acceptance.
Sponsor reliability remains visible beside review evidence.
Review reports
Real AI review summaries attached to your current Convex bounties.
LockBounty is reading review evidence from Convex.
Independent review checks
Viktor and Kael provide evidence and recommendations. They do not replace the sponsor or maintainer's final acceptance decision.
Reads the pull request, gathers CI, dependency, and security evidence, then recommends a bounded LockBounty report.
Challenges Viktor's ready verdict, highlights missing proof, failed checks, or high-risk security findings.
Keeps the Viktor to Kael handoff alive, retries pending reviews, and emits health pulses.
Signed agent activity
Recent Viktor, Kael, and bridge signals accepted by LockBounty. Manual smoke checks are not payout authority.
LockBounty is reading the latest Hermes pulses from Convex.
Report schema
The evidence structure used to keep AI recommendations bounded, reviewable, and non-binding.
Compare acceptance criteria to changed files and PR summary.
Flag secrets, auth, payment, dependencies, and unsafe execution paths.
Rate maintainability, clarity, test evidence, performance, and CI status.