See your traces through a different lens

How the user felt across the conversation, turn by turn, scored by Jev.

A scan of the agent's actions for untrusted input, access to secure data, and external actions.

Where token spend actually went, itemized and reconciled against the model's own numbers.

Review which skills loaded, where each entered the run, and exactly what the agent was told.

Every edit pulled back into a proper code review with a navigable file tree and syntax-highlighted diffs.