What problem does it solve? Plain-language feature requests arrive without impact analysis, proper sizing, or agreed edge cases, leading to mis-sized specs and guessed acceptance criteria. This Skill turns a layman requirement into a verified impact analysis, a decomposition into right-sized feature slices, and Gherkin feature files whose scenarios were confirmed by the user rather than inferred. ## Core Features & Use Cases - Mode A — Decompose: Runs a six-phase gated pipeline (intake, impact analysis, vertical-slice split, sizing, numbering, approval) that writes a requirements document to doc/requirements/ and stops. - Mode B — Author: Brainstorms one feature per invocation using a ten-category edge-case taxonomy (empty state, boundaries, invalid input, duplicates, lifecycle, persistence failure, navigation, offline/time, presentation, multi-actor), then writes features/NNN.{slug}.md and updates all index files. - Use Case: A stakeholder says "users should be able to split a bill with friends." Mode A analyzes the codebase impact and splits it into slices like "record who a bill is split with" and "see what each person owes"; separate Mode B runs then produce user-confirmed Gherkin feature files for each slice. ## Quick Start Ask the AI to run feature discovery on the requirement "users should be able to split a bill with friends" to produce an impact analysis and feature breakdown.
