How kits are tested

A kit is not in stock until it has a job sentence an agent can match, a folder, eval fixtures with a date and environment, a price in USD, and success criteria with a 14-day refund if those criteria fail.

Aaron tests on his own work first — a Raspberry Pi OpenClaw setup, a weekday chief-of-staff brief, workshops for people who do not write code. No brand deals. No invented scores.

Star ratings do not appear on this site. Evals do.