SDET Workflow Framework
Structured delivery, from story to PR.
Three pipelines that turn Claude Code into a systematic engineering partner. Get stories sprint-ready before dev starts, execute structured delivery workflows, then verify with multi-layer QE test execution — with every decision traced and every phase verified.
Three pipelines. One framework.
Story readiness ensures stories are fully specified. Development workflows deliver code. QE workflows verify everything across 6 test layers with failure classification and Jira publishing.
Pipeline 1 — Pre-Development
Story Readiness
Three workflows for PO, Dev, and QE — each adds a layer of specification to the Jira ticket until it's sprint-ready.
Pipeline 2 — Development
Development Workflows
Three structured workflows — bug-fix, enhance, and feature-build — each a complete state machine from ticket to tested PR.
Pipeline 3 — Verification
QE Workflows
Standalone test execution across 6 layers — creates or reuses tests, classifies failures against baseline, publishes results to Jira.
How they connect
Stories flow through readiness into development, then QE verifies. Each stage builds on the previous one.
Each pipeline inherits context from the previous — ACs, architecture, test plans, and code changes flow forward through the full lifecycle.
Up and running in three steps
The framework adapts to your repo — language, test runner, build tools, Jira, feature flags — no manual configuration.
Copy the framework
Drop the .claude folder into your repo root.
cp -r .claude /path/to/your-repo/Generate workflows
Each generator scans your repo and produces tailored workflow files.
/generate-development-workflows/generate-story-workflows/generate-qe-workflowsRun a workflow
Spec stories, ship code, then verify with QE.
/story-draft "Build recommendation engine"/sdet "Login button returns 500"/qe-run PROJ-1234Built on two core ideas
Every workflow — story readiness and development — is powered by the same architecture. Learn it once, understand them all.
LISA
Layered Information State Architecture
Three layers of information: State File (what phase are we in?), Phase Documents (what did each phase produce?), and Knowledge Stores (what did we learn across workflows?).
Learn more →Ralph Loops
Idempotent retry pattern
Every phase checks if it's already done before executing. On failure: retry up to 3 times. After 3 failures: stop and ask for human help. Crashes and interruptions are always safe.
Learn more →
Lisa plans (layered state). Ralph does (idempotent execution).
Quality built in, not bolted on
All three pipelines enforce quality at every stage — from story specification through code delivery and QE verification.
Story Readiness
Definition of Ready
STORY_DRAFTEDStories validated against your team's DoR with structured ACs, scope boundaries, and edge cases.
Technical Refinement
STORY_ENGINEEREDArchitecture, affected components, API contracts, and implementation sub-tasks added by developers.
Test Planning
SPRINT_READY5-layer test plan with coverage targets and test data requirements before a line of code is written.
Development Workflows
Ticket Gate
BLOCKED_NEEDS_HUMANValidates Jira tickets have the right info before discovery starts. Workflow-specific checklists.
Blast Radius
SCOPE_ESCALATIONMapped before changes, verified after. No unintended side effects slip through.
Feature Flags
FEATURE_SHIPPEDNew code behind flags (default OFF). Tested in both states before shipping.
QE Workflows
6-Layer Execution
QE_RUN_COMPLETERuns functional, integration, E2E, performance, security, and accessibility tests — each layer independent.
Failure Classification
REGRESSIONS_FOUNDEvery failure classified as regression, new, pre-existing, flaky, or environment — no manual triage.
Re-Run Intelligence
SETUP_COMPLETEDetects existing tagged tests and previous runs. Reuses what exists, only creates what's missing.
Ready to try it?
Copy the framework, generate workflows for your repo, and run your first structured workflow in minutes.