What problem does it solve? Vague requests like "fix the forms" or "check everything" get claimed done on the wrong evidence, forcing two or three challenge rounds before the real work starts. This Skill converts any ambiguous ask into a five-line brief — objective, definition of done with live evidence, scope boundaries, execution shape, and stated assumptions — so the work matches the intent on the first pass. ## Core Features & Use Cases - Sharpened brief generation: Produces a five-line brief (Objective, Done when, Not doing, Shape, Assumed) and starts work immediately without blocking on confirmation. - Evidence-based done criteria: Forces the "Done when" line to name pasteable, live evidence (probe URLs, test runner summaries, rendered output) instead of "tests pass" or "it works". - Execution shape selection: Routes work into inline, report-only subagent, checkpointed batch, or codify-it shapes based on the request type. - Use Case: A developer says "add validation to the forms". The Skill rewrites it as a brief specifying the full form inventory with a count, a rejected submission shown on screen, edge cases pasted, and the assumption that only user-submittable forms are in scope — then proceeds. ## Quick Start Ask the assistant to sharpen this request before starting: "audit all the endpoints and fix the slow page".
