What problem does it solve? Raw task descriptions are full of blind spots: unstated conventions, hidden constraints, and edge cases the user never thought to mention. This Skill surfaces those unknowns before any work begins, so the final prompt reflects the real task instead of the half-remembered one. ## Core Features & Use Cases - Blindspot pass: Explores the codebase or domain first to find existing modules, conflicting constraints, and edge cases the raw prompt silently ignores. - Structured interview: Asks up to five prioritized questions, preferring recognition-style options over open questions, and skips anything resolvable from the code. - Calibrated prompt output: Produces a ready-to-paste prompt with context, goal, references, constraints, explicit degrees of freedom, a plan-first instruction, and verification criteria. - Use Case: You type /prompt-prep add caching to the API. The Skill scans the repo, finds an existing cache layer and a migration constraint, asks two scoping questions, and returns a refined prompt that pins the decisions and leaves the rest to the model. ## Quick Start Invoke /prompt-prep followed by your raw task description to get a refined, ready-to-paste prompt.
