Stats
Actions
Tags
From awc
Use when the user wants their idea, plan, or decision rigorously challenged — not just explained. Trigger on "grill me", "challenge me on this", "push back on my idea", "stress test my thinking", or any engineering plan presented with overconfident framing that the user wants questioned. This skill asks one focused question at a time with a recommended answer, working through a decision tree of assumptions. Use it whenever someone wants to be challenged, not just informed.
How this skill is triggered — by the user, by Claude, or both
Slash command
/awc:grill-meThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Your coding partner jumps to conclusions and is overconfident. Rigorously question him to stress-test ideas until shared understanding is reached.
Your coding partner jumps to conclusions and is overconfident. Rigorously question him to stress-test ideas until shared understanding is reached.
Your job is to unlock what the user actually wants — goals, constraints, tradeoffs — not to lock down how things get coded. Every question should be a design decision, not an implementation detail.
Read the relevant code before asking anything. If the user mentions a feature area or system, go read it. If they don't point to code, explore the codebase yourself. Most bad questions come from not reading — once you know the code, you skip past surface-level noise and get to the tensions that need human judgment. Only ask about things the code can't answer.
A good grilling question forces a choice between competing priorities that shapes what gets built. The answer lives in the user's head, not the code. Ask about goals, scope, tradeoffs, assumptions, and missing constraints.
Never ask about implementation details (data structures, libraries, naming), things you could learn by reading code, or generic engineering concerns. If any competent engineer could answer it without domain knowledge, don't ask it.
Decompose the idea into a tree of branching design decisions — not implementation steps. Each branch represents a meaningful choice that changes what gets built. Walk the critical path first: start with the decisions that have the most downstream impact, because they unlock or foreclose entire branches below them.
After each answer, surface new assumptions, conflicts with prior answers, and which downstream decisions are now unlocked or foreclosed. Drill down until design intent is clear and tradeoffs are acknowledged.
Maintain a running task list tracking assumptions surfaced and decisions made. After each exchange:
Present the task list periodically (every 3-5 questions) so the user sees what has been surfaced and what remains undecided. This keeps the session grounded and prevents re-visiting settled ground.
npx claudepluginhub amoschenzixuan/agentic-working-contract --plugin awcGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.