Chaos Experiment Plan
Test resilience safely with a small experiment.
Added October 15, 2026. Written with AI assistance and checked for clarity and placeholders; not reviewed by a lawyer, accountant, clinician or other professional. How these prompts are made
The Open buttons send the prompt text as currently shown below to that site only when you click. Fill in the placeholders first.
Plan a safe chaos experiment for [SYSTEM_NAME]. Architecture and critical dependencies: [ARCHITECTURE]. Monitoring in place: [MONITORING]. Environment for the experiment (staging, production with limits): [ENVIRONMENT]. Write an experiment sheet: hypothesis in the form "when X fails, users still see Y"; steady-state measures; the fault to inject and its blast radius; abort conditions with numeric thresholds; communication plan including who must be told; time window; roles; rollback steps; and how to record results. Start with the smallest meaningful fault. Provide a pre-flight checklist and a post-experiment review template. Include three candidate experiments ranked by learning value versus risk. Do not run this in production without approval from the owners and an agreed abort path. Add a maturity path that begins with a game day in staging, moves to a limited production experiment, and ends with regular automated experiments, with the entry criteria for each stage. Include a list of five things that make an experiment unsafe to run (no rollback, unclear ownership, ongoing incident) and a template to record findings and fixes.
Fill in before use
Optional: type your details and the prompt above updates. Nothing is saved or sent.
Replace each bracketed placeholder with your own details, then paste into Any AI or another assistant.
See an example result (fictional data)
Hypothesis: when one cache node fails, users still see pages within 2 seconds. Steady state: p95 latency 600 ms, error rate under 0.5%. Fault: stop one of three cache nodes for 5 minutes in staging, then 1% of production traffic. Abort: error rate over 2% for 60 seconds or p95 over 3 seconds. Roles: driver, observer, communicator. Rollback: restart the node; shift traffic away. Pre-flight: owners informed, alerts silenced only for the test, dashboard open. Candidate experiments ranked: cache node loss (high learning, low risk), slow dependency (high learning, medium risk), region loss (high risk).
Illustrative only, made-up data. Your results will differ and must be reviewed.