grill-me

Stress-tests plans through batched challenge questions about assumptions and failure modes.

Updated Dec 21, 2025
One-click install
npx skills add https://github.com/Angael/veles --skill grill-me-angael
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: grill-me
Source: https://github.com/Angael/veles/tree/main/.agents/skills/grill-me
Command: npx skills add https://github.com/Angael/veles --skill grill-me-angael

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Plans often hide untested assumptions, unresolved contradictions, and overlooked failure modes. This Skill interviews you with targeted, high-leverage questions until the consequential decisions in your plan are explicit and defensible. ## Core Features & Use Cases - Batched Interrogation: Asks 3-7 concise, numbered questions per round, each with a recommended answer and reasoning, so you can respond efficiently by question number. - Codebase-Grounded Questioning: Inspects the repository first and answers verifiable questions itself, reserving your time for genuinely consequential decisions. - Adaptive Follow-Ups: Runs at most three batches, stopping early when the plan is settled, and finishes with agreed direction, assumptions, unresolved risks, and next actions. - Use Case: Before implementing a database migration strategy in a monorepo, get grilled on rollback plans, deployment ordering, and failure scenarios so gaps surface before code is written. ## Quick Start Ask the assistant to grill you on your implementation plan for the current feature before you start coding.

Frequently Asked Questions about grill-me

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I stress-test a project plan before implementation?

Ask to be grilled on your plan and the Skill will inspect the codebase, then pose 3-7 numbered questions targeting consequential decisions and failure modes. Answer by question number, and it adapts follow-up batches based on gaps in your responses.

How many questions does a plan review session ask?

Each batch contains 3-7 concise, numbered questions, with at most three batches total. The session stops early once the consequential decisions are settled, and low-impact details are resolved with stated assumptions.

Does the grilling skill answer questions itself or only ask them?

It inspects the codebase first and answers repository-verifiable questions on its own. Only genuinely consequential decisions that require your judgment are posed as questions, each with a recommended answer and short reasoning.

When should I not use an adversarial plan review?

Skip it for trivial, low-risk tasks where the cost of being wrong is minimal, since the process is designed for consequential decisions. It also does not implement anything; it only clarifies direction, assumptions, and risks unless you explicitly ask for implementation.

What do I get at the end of a grilling session?

You receive a summary of the agreed direction, the assumptions made for low-impact details, any unresolved risks, and the recommended next action. No code is written unless you explicitly request implementation afterward.