An agentic pipeline that bootstraps UI test coverage for web apps by extracting user flows from documentation, the live app, and the codebase, then generating Playwright tests from a structured Flow Map.
User Flow: An atomic, intent-oriented unit describing a single user goal — from a defined precondition to an observable outcome. Described in user-intent terms (not UI/selector terms) so it survives UI changes. Avoid: test case, scenario, user story, journey
Journey: A named composition of User Flows representing a higher-level user goal (e.g. "full onboarding" = sign up → verify email → create first project). Journeys live in the Flow Map; they are not Flows themselves. Avoid: workflow, sequence
Flow Map: The pipeline's universal output artifact. Contains the catalog of User Flows and the Journey compositions that link them. Serves as both human-readable documentation (for non-developer personas) and the structured input to test generation. The Flow Map is the coverage definition — every Flow in the Map gets a test; nothing outside it does. Avoid: test plan, sitemap, spec
Branch:
A named error or edge path attached to a specific step within a User Flow. Each Branch has its own outcome and generates its own test() block, but lives inside the parent Flow in the data model.
Avoid: variant, alternate path, error case
Capability: A shared prerequisite that unblocks test generation for one or more Flows — e.g. an authentication strategy, a test-data fixture, a feature-flag toggle. Capabilities are implemented as code (helpers / fixtures), not as a separate documentation artifact. Flows reference Capabilities by name in their preconditions. Avoid: dependency, helper, prerequisite
Priority: A coarse rank — high / medium / low — assigned to each User Flow. Discovery proposes an initial Priority; the interview phase corrects it. Test generation processes high before medium before low, ensuring high-value coverage lands first regardless of how many Flows the Flow Map ultimately contains. Avoid: severity, importance, rank
Provenance: The set of sources that contributed a User Flow to the Flow Map — e.g. "extracted from README", "discovered via browser exploration", "implied by codebase route", "stated by user in interview". Provenance is recorded per Flow and drives reconciliation: a Flow with weak provenance ("only one source") is flagged for user confirmation; a Flow with conflicting provenance ("docs say X, codebase says Y") is flagged for resolution. Avoid: source, origin, evidence