grill-plan

What problem does it solve?

It reduces the risk of adopting an unclear or weak software/product plan by forcing the draft to be questioned until shared understanding is achieved.

Core Features & Use Cases

• One-question-at-a-time grilling: Resolves the decision tree sequentially to avoid missing critical uncertainties.
• User-answer alignment with a recommended interpretation: Leads each question with the LLM’s best inferred answer from the draft, then confirms or corrects.
• Codebase-first clarification: Prefers checking existing repo knowledge before asking the user when the answer is already discoverable.
• Load-bearing probing and draft edits: Presses on assumptions, ambiguities, success criteria quality, and scope drift, while updating the plan in conversation memory.
• Supports multiple plan sources: Handles both in-context minerva drafts and ad-hoc user-provided plans (e.g., a referenced proposal).

Quick Start

Use the grill-plan skill when you have a drafted plan and want an AI to interview you step-by-step to surface gaps and align everyone before approval.