chaos-experiment-designer
Design DSDS chaos experiments with deterministic setup, explicit hypotheses, and measurable outcomes. Use when planning failure drills, resilience validation, pre-production risk exercises, and repeat
SKILL.mdskills/chaos-experiment-designer/SKILL.md
GitHubChaos Experiment Designer
Overview
Create focused chaos experiments that test one resilience claim at a time. Ensure each experiment is reproducible and has clear pass/fail criteria.
Workflow
- Define the hypothesis
- State one reliability claim in measurable form.
- Identify target metrics and acceptable bounds.
- Define steady state
- Specify baseline workload and expected healthy metric band.
- Establish pre-fault observation window.
- Design fault injection
- Pick minimal fault set to test the hypothesis.
- Set trigger timing and fault duration explicitly.
- Define safeguards
- Add blast radius limits and abort conditions.
- Identify critical user journeys that must stay protected.
- Define evaluation plan
- Add pass/fail rules tied to SLOs and invariants.
- Include recovery expectations (
MTTR, error decay, queue normalization).
- Package experiment
- Return experiment spec, run steps, success criteria, and follow-up checks.
Output format
HypothesisSteady-state definitionFault planSafety and abort rulesPass/fail criteriaFollow-up actions
References
- Read
references/chaos-experiment-template.mdfor reusable experiment skeletons. - Use these repo sources:
docs/05-devs-chaos-and-analysis.md(Chapter 27)canonical-catalogue/DSDS Canonical Catalogue - scenarios to ship.csvcanonical-catalogue/DSDS Canonical Catalogue - Failure modes & propagation semantics.csv