What problem does it solve?
It turns an epic’s locked inputs into implementation-ready, human-approved design artifacts (LLD, verification scenarios, and an execution plan) by fully characterizing code impact and test coverage first.
Core Features & Use Cases
- Architecture-guided LLD generation: Produces interface-level design (tech.yaml) including API contracts, internal interfaces, frontend/backend specs, data specs, communication channels, and mock strategy—explicitly aligned to the locked architecture output.
- Test-driven blast radius and baseline characterization: Maps test surface, computes change surface impact, identifies coverage gaps, and specifies baseline tests capturing current behavior before any design is generated.
- Three-tier verification scenarios + executable task DAG: Creates scenarios.yaml (baseline/new/regression tiers) and plan.yaml (feature slices with depends_on chains and observable exit gates) with strict audience separation.
- Guardrailed, checkpointed locking: Requires explicit human approval (Tether) at context assembly, blast radius, LLD, and scenario/plan stages; locks artifacts only after validation and resolution of open questions/high risks.
Quick Start
Run the prepare play for your epic so it produces locked tech.yaml, scenarios.yaml, and plan.yaml under {stm_base}/{issue}/context/design/.