Grill-Me: Stress-Test Your Plans with a Relentless Agent Interview

Where this fits

You've sketched a feature plan, chosen an architecture, or drafted a go-to-market strategy—but you instinctively know there are holes. Instead of staring at the whiteboard, you want an agent that plays devil's advocate, probing every assumption until the weak spots surface. That's exactly the gap grill-me fills: it turns your AI into a structured interviewer that forces you to defend your thinking, decision by decision.

Why agents benefit

• Systematic frontier expansion: The agent asks only the questions whose prerequisites are already settled, avoiding premature or irrelevant queries—just like a sharp senior engineer would.
• Design tree mapping: Every answer branches into new decisions, so the conversation naturally follows the dependency graph of your plan, not a random Q&A.
• Recommended answers: Each question comes with the agent's own suggestion, giving you a concrete baseline to react to instead of open-ended prompts.
• Relentless persistence: It doesn't let you off the hook—it keeps drilling until you reach a shared, unambiguous understanding of the entire plan.
• No model invocation needed: The skill is designed to be called explicitly, so you control when the interrogation starts, making it predictable and safe to embed in workflows.

Practical scenarios

Product roadmap review

You have a roadmap with three major features. Grill-me forces you to articulate the dependency between them, the metrics for success, and what happens if one slips—revealing that your second feature actually blocks the third, which you hadn't considered.

Database schema design

Before writing migrations, you ask the agent to grill you on your proposed schema. It asks about cardinality, nullability, and query patterns, exposing that your users table needs a tenant_id for multi-tenancy—something you'd missed entirely.

Go-to-market strategy

You're launching a new API. The skill asks about pricing tiers, target segments, and competitive differentiation, pushing you to realize your pricing model contradicts your positioning. The interview sharpens your pitch before you talk to customers.

Add it to your agent workflow

Install the skill with:

npx skills add mattpocock/skills --skill grill-me

Then trigger it in your agent by saying something like: "Grill me on my plan to refactor the auth service." The agent will start the interview, numbering each question and offering a recommended answer. You respond to each round, and it builds the design tree until you both agree the plan is solid.

Compare before adopting

Before you commit, compare grill-me with other research or planning skills. Look at the quality of the questions it generates—are they genuinely insightful or just generic? Check maintenance freshness: does the repo get regular updates? Also consider alternatives like a simple prompt-based interview or a skill that generates a SWOT analysis. Finally, assess workflow fit: does the explicit invocation model match how your agent currently operates, or do you prefer skills that auto-trigger?

Why it is worth tracking

Grill-me comes from Matt Pocock's widely-starred skills repository (222k+ stars), which signals strong community trust and active curation. The skill's design—structured rounds, recommended answers, and a clear termination condition—is a pattern worth copying even if you don't adopt it directly. Track it as a reference for building your own interrogation-style skills. When you need to stress-test a plan, this is the first tool to evaluate. For a broader view of similar agent skills, check OpenAgentSkill's listing to see how it compares with other research and planning utilities.