Grill me

Skill sruja-ai/sruja/skills/grill-me

A CLI-first autonomous coding agent where the actor never grades itself. Every edit is grounded in your repo's real topology and verified by an independent deterministic grader (drift, lint, verify-task, intent) before it ships. Closed-loop: comprehend → plan → execute → critique → replan. Also a passive MCP harness.

Install

npx -y skills add sruja-ai/sruja --skill grill-me

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

24 stars24 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.1 KB, as published. Nobody here has run it

Grill Me - Shared Understanding Skill

This skill ensures AI and human share a design concept before writing code.

Description

Use this skill when planning new features or significant changes. The AI will interview you relentlessly to reach shared understanding.

Usage

# Activate skill
@skill:grill-me

# Or use in prompt
"Grill me on this plan until we reach shared understanding."

Process

You describe what you want to build
AI asks questions to clarify:
- What problem does this solve?
- Who is the user?
- What are the success criteria?
- What are the dependencies?
- What could go wrong?
- How do you measure success?
Continue until AI has no more questions
Document the shared understanding as PRD or issues

Questions AI Will Ask

Understanding Phase

What exactly are you trying to build?
Why is this important?
Who will use this?
What happens if we don't build it?

Requirements Phase

What are the acceptance criteria?
What are the edge cases?
What are the error conditions?
What are the performance requirements?

Architecture Phase

What does this depend on?
What will depend on this?
Where does this fit in the architecture?
What are the boundary conditions?

Implementation Phase

How will you verify this works?
What could go wrong in production?
What is the rollback plan?
How do you measure success?

Example

You: I want to add a feature that shows code complexity
AI: What problem does this solve?
You: Developers don't understand their code quality
AI: What specifically do you want to measure?
You: Cyclomatic complexity
AI: What threshold indicates "too complex"?
You: 10
AI: What happens when it exceeds threshold?
AI: [continues until full understanding]

Output

After grilling, produce:

Product Requirements Document (PRD)
Or direct to issues/tasks
Never start coding until shared understanding reached

Why This Matters

From the talks: "Me and the AI don't share a design concept."

Without shared understanding, AI produces garbage
Re-running compiler just produces more garbage
Solution: reach shared understanding first