grill-me

Stress-tests implementation plans through structured adversarial interview rounds before coding.

Updated Sep 10, 2026
One-click install
npx skills add https://github.com/loteiron/ZeusAgent --skill grill-me-loteiron
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: grill-me
Source: https://github.com/loteiron/ZeusAgent/tree/main/optional-skills/software-development/grill-me
Command: npx skills add https://github.com/loteiron/ZeusAgent --skill grill-me-loteiron

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Plans often contain hidden assumptions and unresolved decisions that surface only after code is written, causing rework and misalignment. This Skill interviews you with adversarial, dependency-ordered questions until every design decision is explicitly settled. ## Core Features & Use Cases - Frontier Rounds: Models your plan as a design tree and asks only the questions whose prerequisites are already resolved, each with a recommended answer. - Fact-Gathering Autonomy: Looks up codebase facts itself using file search and terminal tools instead of asking you questions it can answer. - Decision Synthesis: Summarizes all settled decisions, open items, and out-of-scope boundaries, then waits for explicit alignment before implementation. - Use Case: Before building an auth flow or schema migration, say "grill me on this plan" to expose edge cases, failure modes, and rollback strategies you had not considered. ## Quick Start Ask the agent to grill you on your implementation plan for the upcoming feature before writing any code.

Frequently Asked Questions about grill-me

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

FAQPage Schema
What is a design tree in software planning?▼

A design tree maps every plan decision as a node branching into dependent decisions. The frontier is the set of questions whose prerequisites are settled, so they can be asked now without guessing at unanswered dependencies.

When should I use adversarial plan review instead of code review?▼

Use adversarial plan review before any code exists, especially for complex work like auth flows, migrations, or payments. Code review applies to existing implementations; plan interviews catch flawed decisions before they become expensive rewrites.

Can the interview questions be answered from the codebase automatically?▼

Yes. Fact-based questions about the codebase, filesystem, or configuration are resolved by the agent using file search, read, and terminal tools. Only genuine decisions requiring user judgment are asked in interview rounds.

When should I not use plan stress-testing?▼

Avoid it for simple one-off tasks where overhead exceeds value, and for reviewing existing code, where a code review skill is the right tool. It is designed for plans with unresolved decisions or vague scope.