The most interesting thing in @agenticscredit's scoring docs might be when it refuses to give you a score. The current requirement is at least 20 closed trades across 14 active days. Anything thinner can come back as insufficient_data instead of forcing an ACS onto a weak sample. That sounds like a small implementation detail, but it says something bigger about the system. Reputation gets unreliable when a model pretends limited history is enough evidence. Here, we don’t know yet is an actual state. I like that design choice. Crypto dashboards are very good at producing instant numbers. A credible scoring layer also needs to know when it has not seen enough to produce one.