Skip to content
Back to skills

Grilling

ASecurity

Interview the user relentlessly about a plan, decision, or idea until every branch of the decision tree is resolved — one question at a time, each led with a concrete recommendation. Use before large or ambiguous undertakings, when requirements feel underspecified, or on 'grill me', 'stress-test my plan', 'löcher mich', 'hinterfrag meinen plan', 'denk das mit mir durch', 'frag mich aus'.

  • 2 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added September 27, 2026
developmentgoreactgit

Security analysis

A100/100

Scanned September 27, 2026

npx -y skills add emanuelrechsteiner/claude-rcode --skill grilling --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Grilling?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Grilling
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/emanuelrechsteiner-grilling/badge)](https://www.skillsdirectory.com/skills/emanuelrechsteiner-grilling)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: grilling
description: "Interview the user relentlessly about a plan, decision, or idea until every branch of the decision tree is resolved — one question at a time, each led with a concrete recommendation. Use before large or ambiguous undertakings, when requirements feel underspecified, or on 'grill me', 'stress-test my plan', 'löcher mich', 'hinterfrag meinen plan', 'denk das mit mir durch', 'frag mich aus'."
context: main
---

# Grilling

Interview the user relentlessly about every aspect of the plan, decision, or idea until you reach a **shared understanding**. Walk down each branch of the decision tree, resolving dependencies between decisions one by one. The most common failure mode in software work is misalignment — this session closes the gap BEFORE any work begins.

## Rules

1. **One question at a time.** Ask, wait for the answer, then continue. Multiple questions at once are bewildering and produce shallow answers. (This deliberately overrides the batch-questions default — depth beats throughput here.)
2. **Lead every question with a recommendation** plus a one-line situational WHY, per `rules/recommend-on-ask.md`. The user reacts to a concrete proposal faster than to an open question.
3. **Facts vs. decisions.** If a *fact* can be found by exploring the environment (filesystem, git history, tool output), look it up instead of asking. *Decisions* belong to the user — put each one to them and wait for the answer.
4. **Do not act** on the plan until the user confirms the shared understanding is reached.
5. **Close with a decision record.** When the tree is resolved, output a compact summary of every decision taken (and rejected alternatives where meaningful). Feed it into the planning doc (`rules/planning-doc-convention.md`) and, where terms or architectural choices were coined, into `CONTEXT.md` / an ADR (`rules/domain-docs-convention.md`).

## When NOT to use

- The task is small and unambiguous — grilling overhead exceeds its value.
- The user already provided a complete spec — go straight to work.
- Mid-implementation detail questions — ask those inline, one-off.

---
*Adapted from [mattpocock/skills](https://github.com/mattpocock/skills) `grilling` (MIT License, © 2026 Matt Pocock); integrated with this framework's recommend-on-ask, planning-doc, and domain-docs conventions (IMP-124).*

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…