Feature Forge makes Claude act as a requirements specialist. It discovers the goal and users, interviews you with structured choices and open questions, writes functional requirements in EARS syntax, validates the acceptance criteria with you and ends with an implementation checklist.
Five reference files hold the EARS patterns, interview questions, a specification template, acceptance criteria examples and a pre-discovery pattern that uses subagents for features that span several domains.
What You Get
- A specification with overview, user value and functional requirements in EARS format
- Non-functional requirements for performance, security and scalability
- Acceptance criteria in Given/When/Then form
- An error handling table with status codes and messages
- An implementation TODO checklist split by backend, frontend and testing
Common Use Cases
- Defining a new feature before any code is written
- Turning a rough product request into testable requirements
- Writing a PRD with acceptance criteria for a team
- Gathering context across several system layers before the interview
What Makes It Unique
The interview comes first: the skill forbids generating a spec without one and prefers structured choice questions for priority, scope and format. Requirements use EARS patterns (ubiquitous, event-driven, state-driven, conditional and optional), and acceptance criteria are checked against INVEST. For multi-domain features it can launch parallel subagents to read the codebase before the interview starts.
Under The Hood
The skill is a SKILL.md with five reference files and no scripts, and the references load by topic. It tells the agent to ask you questions through the AskUserQuestions tool and, for features that span several domains, to start Task subagents that explore the codebase and apply other skills. The finished specification is written to a file named specs/{feature_name}.spec.md in your project. The skill does not run shell commands, fetch web pages or call outside APIs.