What problem does it solve? Product teams often hand engineers feature descriptions that are too vague to build or test against, causing rework, scope disputes, and untestable acceptance criteria. This Skill turns the features named in an approved PRD into implementable, verifiable feature specifications with observable behavior, complete I/O contracts, and pass/fail acceptance criteria. ## Core Features & Use Cases - Per-feature elaboration method: Traces every feature back to a PRD line, specifies observable behavior with EARS phrasing and use-case flows (main/alternate/exception), and enumerates inputs, outputs, state-transition tables, and decision tables. - Testable acceptance criteria: Writes Given/When/Then criteria for deterministic features and metric-threshold criteria on named datasets for probabilistic/ML features, plus edge cases each paired with their expected handling. - Archetype overlays and amend workflow: Applies per-feature NFR targets and archetype overlays (UI, REST/API, data-ML, batch, integration, CLI), and supports versioned in-place amendments with bidirectional ripple analysis across PRD, design, and test docs. - Use Case: Given an approved PRD naming an invoice-reminder feature, produce a spec where each behavior has a trigger and observable response, every edge case names its handling, and a tester can verify each criterion without asking the author. ## Quick Start Use authoring-feature-spec to elaborate the features in my approved PRD into a testable feature specification.
