Reliability

When to use

Invoke after system-design has approved a design and before implementation and platform work hardens the runtime. Use it whenever a system has externally meaningful availability commitments, multi-component failure interactions, or stateful dependencies whose loss requires a recovery plan.

Do not use for telemetry pipelines or oncall workflow design (use operations), latency and throughput budgeting (use performance), security threat modeling (use security), or post-incident retrospectives on already-shipped systems.

Inputs

Required:

• Approved system-design.md and its relevant ADRs.
• The reliability scope in question: the user-facing journeys, APIs, and background workflows whose availability must be defined.
• Dependency inventory: upstream services, datastores, third-party APIs, and platform substrates the system relies on.

Optional: