What problem does it solve? After a decision is made, teams often produce vague specs with untestable requirements like "should be fast," leading to misaligned implementations and endless re-litigation of the premise. This Skill enforces a fixed-structure specification with measurable constraints and falsifiable acceptance criteria. ## Core Features & Use Cases - Structured Input Gathering: Collects required slots (artifact, objective, constraints, interfaces, acceptance, evidence base, exclusions) and asks before filling gaps with assumptions. - Fixed Output Format: Produces specs in a defined order — summary, MUST/MAY/MUST NOT requirements, measurable constraints, typed interfaces, falsifiable acceptance criteria, explicit out-of-scope items, and labeled open questions. - Failure-Condition Self-Check: Rewrites untestable requirements, adds missing error-case behavior, and removes exploratory language before emitting the spec. - Use Case: After deciding to build a new API endpoint, invoke the skill to produce a buildable spec with p95 latency thresholds, typed request/response schemas, and testable acceptance criteria an engineer can implement directly. ## Quick Start Ask the AI to run /specification and provide the artifact, objective, constraints, interfaces, and acceptance criteria for the thing you want specified.
