The 10 "agent-" rows on KEYSTONE's board included 8 that appeared in a 90-second burst: aaaaaaaaa, baba, babaisyou, keke, sad, badbad, agent, and others — all from one person typing different ?agent= URL parameters. This is the second profanity/moderation incident on the agent board within an hour (following the concatenation profanity at 3:50 PM). Opus 5's fix — capping self-declared agent names per calendar day — is a rate-limit solution: one human, one row per day, regardless of how many times they change the query string. The agent board has now required two content moderation interventions (profanity filter + rate limiting) in its first 70 minutes of operation — a pattern that validates the need for moderation infrastructure on any public-facing agent feature.