kiro-discovery

To determine the correct action path for a feature request, a discovery workflow classifies whether it should update an existing spec, become a new spec, decompose across multiple specs, or be implemented directly, then surfaces the recommended next action.

Spec-driven discovery is the process of evaluating a proposed feature or bug to decide its implementation scope. You need it when onboarding new ideas into a repo to determine whether to initialize a new spec, update an existing one, or perform direct implementation work.

To decide whether to update an existing spec or create a new one, a lightweight project scan inspects spec manifests and steering file presence, categorizes the project state, and applies path decision logic to classify the request accurately.

Yes, spec-driven discovery works with greenfield projects by performing a metadata-only scan to detect the absence of existing spec folders, which triggers the recommendation to initialize a new spec before proceeding with implementation.

Before proposing approaches and writing spec briefs, the workflow gathers context by loading steering files and adjacent specs, optionally dispatching subagents to explore the codebase, and asking sequential clarifying questions to discover boundaries.

Discovery session artifacts are preserved by writing brief.md and roadmap.md files directly to disk during the workflow, which records the proposed approaches, viability checks, and task decomposition details for continuous roadmap planning.