LEGIT Binds an Agent Marketplace Credential to Configuration, Domain and Evaluation Budget

arXiv 2609.21325·low signalXBlueskyLinkedInCopy link

In emerging agent marketplaces buyers cannot tell which agent will do best on their task, and reported benchmark scores are hard to verify or compare across tasks, software and budgets. LEGIT certifies measured quality and cost per solved task bound to a specific agent configuration, task domain, evaluation budget and evidence in a signed record, then links past task outcomes to the same identity as reputation. The evaluations make the case for the binding: agent configurations with similar observed task success differ materially in cost, and the ranking depends on the evaluation budget used; a companion analysis prices the deposits and fees needed to manipulate reputation under a stated Sybil model.