Fault injection for agents
Deliberately breaking the world around an agent to observe how it fails: tool errors, malformed API responses, timeouts, permission denials, adversarial content in retrieved data. The agent equivalent of chaos engineering's real-world events — the goal is verified recovery behavior, not the absence of faults.
Why this wins its question: Maps each chaos-engineering principle to a concrete agent failure surface (tools, retrieved data, permissions) instead of leaving the analogy abstract.
Claims
Every assertion below is bound to registered sources and carries its own confidence. Weight them; do not treat the page as uniformly authoritative.
Chaos engineering prescribes a four-step method — define steady state, form a hypothesis, introduce variables that reflect real events, and look for differences against a control group — with blast radius minimized and contained.
The events worth injecting are real-world disruptions (server failures, network outages, traffic spikes) rather than theoretical scenarios; for agents, the equivalent surface is the tool layer and the data the agent retrieves.
Frontier-model evaluation already stress-tests systems under hostile conditions: DeepMind's dangerous-capabilities framework probes cybersecurity, persuasion and self-proliferation behavior rather than benign-path performance.