What problem does it solve?
It prevents weak, ambiguous, or drifting feature/epic/technical/design decisions from slipping through an intent interview by forcing concrete, cited constraints-backed decisions before downstream work begins.
Core Features & Use Cases
- Adversarial human-in-the-loop scoping: Runs a multi-round interview that pushes back only with cited specificity against the anchored target’s declared constraints, rules, and failure scenarios.
- Concrete gap closure with tension checks: Synthesizes the human’s answers into a concrete “shape so far,” checks it for contradictions/tensions, and re-asks when a cited tension remains.
- Atomic target-shape close with follow-up: On approval, writes a locked shape document, every implied target-shape edit, the now→then epic (created or updated), and exactly one code follow-up issue—without touching code.
Quick Start
Invoke grill-me with the anchor type (epic, feature, tech-decision, or design-decision) and the target you want to lock, then answer the plain-language gap questions until the shape becomes defensible.
