Agent Effects: test AI agents’ side‑effect reliability with fault injection
Agent Effects provides a CLI and Python library to inject realistic faults into an AI agent’s workflow, capture the final business state, and reduce failures to reproducible regression cases. It is aimed at platform engineers building payment, messaging, or infrastructure agents, giving them a lightweight, no‑API‑key tool to validate idempotency and reconciliation. Unlike generic unit tests, it focuses on durable side‑effects, making it easier to catch double‑processing or lost‑acknowledgment bugs. The tool ships with examples, a pytest plugin, and a static HTML report for CI integration.
View on GitHub →eschmidt01/agent-effects-testkit