# Grill Me

> Get relentlessly interviewed about a plan, design, or change until every branch of the decision tree is resolved — depth-first, one question at a time, recommendation-attached, code-checked before user-asked. Use before any non-trivial change, when scope feels fuzzy, or to stress-test a plan before committing. Trigger on: 'grill me', 'interview me', 'challenge my plan', 'ask me questions', "I'm not sure what I want", 'help me think this through', 'is this the right approach', 'walk me through it', 'stress test this plan'.

- Skill: `ivy00johns/grill-me` (Agent Skill)
- Install (CLI): `npx skillmds@latest add ivy00johns/grill-me`
- Raw SKILL.md: https://api.skillmd.com/api/skills/ivy00johns/grill-me/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: ivy00johns (https://skillmd.com/u/ivy00johns)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/ivy00johns/grill-me

---


# grill-me

Walk the design tree depth-first. Resolve the current branch before backing out. Three rules:

1. **One question at a time.** Wait for the user's response before asking the next. Don't batch.
2. **Recommend, then ask.** Every question takes the form: "I think *<X>* because *<reason>*. Do you agree, or do you want *<alternative>*?" The recommendation is the default — the user accepts with "yes" or steers with an alternative.
3. **Ask code, not user, when possible.** Before asking, check: can `Read`, `Grep`, or `Glob` answer this? If yes, do that and proceed. Only ask the user when the answer is genuinely in their head.

**Exit condition:** every decision in the tree has a chosen answer plus your analysis attached. Output a final `Decisions` section the user can paste into a plan or brief — one bullet per decision, with the chosen answer and a one-line rationale.

**Compose with:** `architecture-rescue` when grilling about an unfamiliar codebase; `maintain-context` to capture decisions as ADRs or glossary entries; `plan-builder` to turn the `Decisions` output into an executable plan. `find-unknowns` is the sibling for the other direction — this skill interrogates the *known* unknowns you already have opinions on; run its blindspot pass first when you don't yet know what to have opinions about.