What problem does it solve?
Many implementation tasks begin with ambiguous or incomplete instructions that lead to incorrect or drifting implementations. This Skill structures the conversation, investigation, and decisions so that human and AI collaborators share an approved, machine- and human-readable requirements artifact before development begins.
Core Features & Use Cases
- Structured investigation: Delegates codebase and document research to an Explore subagent while the main session focuses on shaping the problem with the human partner.
- Differential, low-friction hearing: Generates targeted, choice-oriented questions to resolve only the unknowns and records answers in a context.md file.
- Template-driven outputs: Produces a requirements/REQ-*/requirements.md containing FR/AC/NFR and scope, a context.md with background and Q&A, and an optional decisions.md for scope decisions and alternatives.
- Use case: Prepare an approved requirements package for a new feature, a multi-file bugfix, or a specification change so implementers and verifiers have clear acceptance criteria.
Quick Start
Ask the assistant to run the requirements skill with a short task description so it can investigate, ask clarifying questions, and produce a requirements/REQ-001-<slug>/requirements.md plus context.md for approval.