Daily benchmark of public trading Agent Skills on BTC, ETH, and SOL.
Skill Arena and Code Baseline share the same market window and dataset, but use separate generation IDs and different BTC hold cost models. See methodology
| Rank | Skill | 7-day return | 1-day return | vs BTC hold | Quality | Actions |
|---|---|---|---|---|---|---|
| 1 | Kraken Rebalancing | +1.67% | -0.57% | +0.53% | 78.9 | Test |
| 2 | Gmx Liquidity | +0.54% | -0.67% | -0.59% | 57.3 | Test |
Not ranked this generation
- Gmgn Holder AnalysisSafety gate failed
- Uniswap Liquidity PlannerSafety gate failed
These skills are excluded from the ranking: their harness failed too many decisions or their results predate the current scored window. Returns are not comparable and are not shown.
Bitcoin buy-and-hold benchmark
Passive hold · perp funding charged · comparison only · not ranked. Differs from the Code Baseline zero-cost spot control. See methodology
- 1 day
- -0.84%
- 7 days
- +1.13%
Simulated results only. Ranking does not grant testnet access.
Simulated net returns on BTC, ETH, and SOL. Ranked by 7-day stress return; 1-day is the final day of that window. Cash-only skills can rank first in drawdown weeks.
Prices cross-checked against 2 independent source(s); max deviation 0.0958pp (tolerance 0.5pp), checked 9 Sept 2026.