grilling

Sharpens loose ideas into documented decisions through structured interview rounds.

Updated Jul 28, 2026
One-click install
npx skills add https://github.com/vinzenz/workflows --skill grilling-vinzenz
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: grilling
Source: https://github.com/vinzenz/workflows/tree/main/.agents/skills/grilling
Command: npx skills add https://github.com/vinzenz/workflows --skill grilling-vinzenz

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Loose ideas and fuzzy feature requests often get built before anyone pins down what they actually mean, leading to rework and misaligned implementations. Grilling runs a relentless interview that challenges vague terms, invents edge cases, and captures every decision as a glossary entry or ADR the moment it crystallises. ## Core Features & Use Cases - Batched question rounds: Asks 4-6 related questions per round with pre-selected recommendations, delivered through galley so the agent blocks until the human answers. - Live documentation: Writes CONTEXT.md glossary entries and numbered ADRs in docs/adr/ as decisions happen, never batched at the end. - Sharpening techniques: Challenges terms against the glossary, stress-tests with invented edge cases, and cross-references claims against actual code. - Use Case: A stakeholder says "add account linking." Grilling runs ~5 interview rounds to pin down what "account" means, which entities link, and what happens on conflicts, producing a CONTEXT.md glossary and ADRs a developer can build from. ## Quick Start Ask the agent to grill you on your new feature idea until every fuzzy term and design decision is pinned down and written into CONTEXT.md and ADRs.

Frequently Asked Questions about grilling

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

FAQPage Schema
How do I turn a vague feature idea into concrete design decisions?▼

Run an interview loop: orient by reading existing code and docs, ask batched rounds of 4-6 questions with pre-selected recommendations, and write each decision into a glossary or ADR immediately. Repeat until no answer changes what gets built.

How to run human-in-the-loop question rounds with galley?▼

Write one self-contained HTML artifact per round to .galley/<round>.html, audit it with galley check, open it with galley open, then block on galley wait until the human answers. End the session with galley end when the interview finishes.

When should I write an ADR for a design decision?▼

Write an ADR only when all three hold: the decision is hard to reverse, it is surprising without context, and it resulted from a real trade-off between genuine alternatives. If any condition is missing, skip the ADR.

What happens when no human is available to answer questions?▼

The round cannot run. Do not emit questions into the void or record your own recommendation as the decision; leave the work unresolved and state that a human-in-the-loop session is needed.

When should I use wayfinder instead of grilling?▼

Use wayfinder when the effort is bigger than one session, such as many interlocking decisions spanning weeks of work. Grilling covers one session's depth; wayfinder handles larger multi-session discovery efforts.