What problem does it solve?
Writing a feature spec that downstream LLM agents can actually implement requires gathering codebase context, market research, user decisions, and technical design in one coherent document — a process that usually produces vague specs, missed edge cases, and rework. This Skill structures that entire authoring process into a two-stage workflow with mandatory research and decision grilling before any file is written.
Core Features & Use Cases
- Two-stage spec authoring: Produces one
_spec.md with a Product part (WHAT, WHY, WHO) and a Technical part (HOW), separated by a user-confirmed checkpoint.
- Companion catalogs: Generates
_user_stories.md with acceptance criteria and edge cases, a _dx.md developer-experience contract, a _uiux.md UI change map for UI-bearing features, and a _tests.md test contract with stable IDs.
- Frontier grill protocol: Runs parallel codebase and market research, then interviews the user in rounds until every design decision is resolved, recording significant trade-offs as ADRs.
- Use Case: A developer wants to add webhook notifications to their product. The Skill researches the codebase and competing products, grills them on which events should notify users, freezes the CLI/API surface, then writes the full spec, story catalog, and test contract ready for task breakdown.
Quick Start
Ask the agent to create a spec for your new feature idea, providing the feature name and any existing notes you have.