Przejdź do treści
Biblioteka skilli
Zarządzanie i produkt · Planowanie

grilling

/grilling

Przepytuje użytkownika w rundach, aż każda decyzja, na której opiera się plan, specyfikacja lub mapa, zostanie rozstrzygnięta.

Repozytorium
Claude-Quorum ↗
Licencja
MIT
Ostatnia zmiana repozytorium
12.11.2026

Ocena i bezpieczeństwo

100/ 100Klasa: dobry
Format25 / 25
Frontmatter z nazwą i opisem, poprawna nazwa, opis do 1024 znaków, nazwa zgodna z folderem.
Odkrywalność25 / 25
Opis mówi, kiedy użyć skilla, ma sensowną długość i jest w 3. osobie.
Treść35 / 35
Co najmniej 150 słów i do 500 linii, nagłówki, listy kroków, przykłady albo kod.
Higiena15 / 15
Bez zaślepek typu TODO, bez ścieżek Windows, z licencją.
Ryzykobrak sygnałów w skanie
Pliki w paczce
1
Skrypty
0
Komentarze HTML
0
Adresy URL w SKILL.md
0
Przestarzałość

Brak sygnałów. Repozytorium zmienione 12.11.2026, bez odwołań do wycofanych modeli i API.

Licencja

MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.

Co robi i kiedy

Przepytuje użytkownika w rundach, aż każda decyzja, na której opiera się plan, specyfikacja lub mapa, zostanie rozstrzygnięta.

Oryginalny opis (EN)

Quorum's grilling habit. Interview the user in rounds until every decision a plan, spec, or map rests on is settled. Each round asks every question that can be asked now, numbered, each with a recommended answer; facts get looked up, never asked. Use when the user asks to be grilled or to stress-test a plan or idea, and wherever a Quorum step says to grill. It only asks and records answers; it never starts a Quorum command.

Jak działa

  1. The design tree
  2. A round
  3. Rules
  4. Ending
Sekcje z nagłówków SKILL.md, w oryginale · 60 linii, 622 słów
Podgląd SKILL.md
---
name: grilling
description: Quorum's grilling habit. Interview the user in rounds until every decision a plan, spec, or map rests on is settled. Each round asks every question that can be asked now, numbered, each with a recommended answer; facts get looked up, never asked. Use when the user asks to be grilled or to stress-test a plan or idea, and wherever a Quorum step says to grill. It only asks and records answers; it never starts a Quorum command.
---

# Grilling

A plan goes wrong where it rests on a decision nobody made: something the agent assumed, or a question asked so
loosely that the answer didn't settle it. Grilling finds those decisions and gets the user to make each one, in as
few rounds as the questions allow.

The step that sends you here says what's being decided, where answers go, and when to stop. Without one, you're
grilling a plan or idea the user brought; record the decisions in your reply.

## The design tree

Each decision opens others: choose "bill per seat" and "what counts as a seat" follows. Questions whose
prerequisites are settled are the **frontier**: you can ask them now without guessing at answers you haven't heard.

- **Depth first** (the default): follow the frontier until each branch reaches something concrete enough to build
  or check.
- **Breadth first** (when the step asks, for example charting a map): one level across every branch before going
  deeper, to see the whole shape and the questions that can't be asked yet.

A question that depends on another one still open belongs in a later round, not this one.

## A round

Ask the whole frontier at once, in one message. With AskUserQuestion, that's up to 4 questions per call; if the
frontier holds more, ask the 4 whose answers unblock the most, and the rest next round. Without it, number them in
one message. Each question:

- **A short title**, and one or two sentences: what's being decided and why it matters here. No more. Long
  questions tire people and hide why you're asking.
- **The choices**, when there are a few, each in a few words.
- **Your recommended answer**, first (with AskUserQuestion, label it "(Recommended)"), and one line on why. Word the
  question so "yes" accepts it.

Then wait. Each round's answers reshape the tree: work out the new frontier before the next round.

## Rules

- **Facts are yours to find.** Never ask what the code, the docs, `CLAUDE.md`, or a command could tell you. Look it
  up, or start an `Explore` agent for a wide search, and ask only about intent, priorities, scope, and trade-offs. A
  lookup still running blocks only the questions that wait on it.
- **Don't ask twice.** Before each round, check what's already decided: the request, earlier answers, the spec,
  `research.md`, a map's decisions (`decisions.md`), `REVIEW.md`. Cite a settled answer instead of asking again. If
  it now looks wrong, say why and ask whether to change it.
- **The user's answers are the user's.** Never answer your own question, and never take silence as a yes. If they
  say "you decide", record "the user left it to me: <your recommendation>".
- **Challenge a vague answer.** "Fast", "secure", or "the usual way" isn't a decision. Ask for the number, the case,
  or the example that would settle it, once.
- **Name the trade-off you heard.** If two answers pull against each other, say so and ask which wins.
- **With `--yes`** (only when the user typed it), don't ask: take each recommended answer, and list them as
  assumptions in your reply, except where the step says the user must answer.
- **Most specs need one round, sometimes two.** If you're on a fourth round, the scope is probably too big: say so,
  and suggest splitting it (for an effort bigger than one task, `/quorum:map`).

## Ending

Stop when the frontier is empty: every branch either decided, ruled out of scope, or left open on purpose as a known
gap. Then list the decisions in a few lines and ask whether anything's wrong. Don't act on them before the user
confirms, unless the step that sent you here says what comes next.

Wersje

RepozytoriumWynikLicencja
Ta strona · SimplySylvia/Claude-Quorum100MIT
nikospavlopoulos/devops-mutant-powers100MIT
RIJUDEVx/agent-skills100MIT
yuan88964/RelayStudio100MIT
FrancoStino/opencode-skills-collection100MIT
cravingmaker/skills100MIT
helmholtzai-consultants-munich/helmholtz-agent-skills100Apache-2.0
bendrucker/claude100MIT
timharris707/skills100MIT

Migawka bazy z 13.11.2026 · wersja paczki 268624fcbd76 (skrót SHA-256 całej paczki)

Pobrania w ostatnich 30 dniach: 0 · ranking

Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.

Podobne skille

  • 84grill-me
    Przepytuje użytkownika w rundach, aż każda decyzja w planie zostanie rozstrzygnięta, a fakty sprawdza samodzielnie.
  • 100ra-grilling
    Dopytuje użytkownika o plan, decyzję lub architekturę, aż każda gałąź drzewa projektu zostanie rozstrzygnięta.
  • 100grill
    Przepytuje osobę z planu lub decyzji w rundach numerowanych pytań, aż każda gałąź zostanie rozstrzygnięta.
  • 84sw-grill-me
    Przepytuje użytkownika w rundach, aż plan będzie bez otwartych decyzji, a fakty sprawdza sam.
  • 100plan-grilling
    Przepytuje plan, projekt lub pomysł w kolejnych rundach pytań, aż wszystkie decyzje zostaną rozstrzygnięte przed budową.
  • 90grill-designs
    Przepytuje użytkownika decyzja po decyzji, aby przetestować plan lub projekt, aż cele, ograniczenia, założenia, kompromisy, ryzyka i weryfikacja staną się jawne.

Uwagi i zgłoszenia

Wiadomość trafi do Bartosza Sokolińskiego. Adres e-mail jest potrzebny tylko do odpowiedzi. Szczegóły w polityce prywatności.