← ClaudeAtlas

grilllisted

Interrogate a plan, design, or decision one question at a time until it holds up, capturing the agreed vocabulary in a glossary and each hard call as a written decision as they settle. Use when the user asks to grill, stress-test, or pressure-test their thinking; not for routine implementation work or questions that have a single correct answer.
wizspec/agent-workflows · ★ 3 · AI & Automation · score 79
Install: claude install-skill wizspec/agent-workflows
# Grill Interview the user relentlessly about a plan, design, or decision until you reach a shared understanding, and write down the language and the decisions the moment they crystallise. The interview and the documentation are one activity. Grilling is what surfaces the vocabulary and the trade-offs; capturing them immediately is what stops the same ground being retrodden next week. Writing up afterwards loses the precision that the questioning just bought. Do not act on the plan until the user confirms you have reached a shared understanding. ## Read the model before the first question Find the project's existing domain model and read it before opening the interview: - a `CONTEXT-MAP.md` at the repository root means the project has several contexts, and it records where each one lives; - otherwise a root `CONTEXT.md` means a single context; and - neither means nothing has been written down yet. Existing decisions usually live in `docs/wizspec/decisions/`, with context-specific decisions beside the context they belong to. Read what is already there. You cannot challenge a term as inconsistent without knowing what the project has already agreed it means. ## Run the interview Ask one question at a time and wait for the answer before moving on. Asking several at once is bewildering and produces shallow replies. Walk down each branch of the decision tree, resolving dependencies between decisions one by one. Take the decisions that constrain later ones first. For ev