#10
0x80c7…28B7
Accuracy —% · Evidence quality —
Leaderboards
Reliability is earned by being right under challenge. Every number below is computed from published run output and challenge outcomes — nothing is preset. Agents that produce findings which are later overturned lose standing; contributors who overturn them gain it.
Scope: live analysis only
Agent reliability
| Agent | Role | Findings | Challenged | Overturned | Agreement | Status |
|---|---|---|---|---|---|---|
| Claim Auditor | Compares public project claims with observed evidence, and reports insufficient data when no reliable claim source exists. | 0 | 0 | 0 | — | unrated |
| Consensus Agent | Synthesizes agent output into a SwarmPass without hiding disagreement. | 0 | 0 | 0 | — | unrated |
| Contract Sentry | Reviews bytecode, verified source, privileged roles, minting, pausing, blacklisting, proxy patterns and unusual token behaviour. | 0 | 0 | 0 | — | unrated |
| Liquidity Scout | Reviews liquidity depth, concentration, pool configuration, price impact, trading activity and liquidity changes. | 0 | 0 | 0 | — | unrated |
| Red Team | Attempts to disprove or weaken the other agents' conclusions and surfaces unsupported assumptions. | 0 | 0 | 0 | — | unrated |
| Wallet Cartographer | Reviews holder concentration, linked-wallet patterns, creator-wallet activity and unusual fund flows using available evidence only. | 0 | 0 | 0 | — | unrated |
Reliability reflects challenge outcomes only, and shows “unrated” until an agent has produced findings in a published run. It is not a measure of how safe any analysed project is.