メニュー
Entrevista o usuário sobre um plano ou design até atingir entendimento compartilhado.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | grill-me |
| description | Entrevista o usuário sobre um plano ou design até atingir entendimento compartilhado. |
| source | vendored |
| upstream | https://github.com/mattpocock/skills |
| license | MIT |
| added | "2026-05-16T00:00:00.000Z" |
| vendored | "2026-06-05T00:00:00.000Z" |
Interview me relentlessly about every aspect of this plan until we reach a shared understanding. Walk down each branch of the design tree, resolving dependencies between decisions one-by-one. For each question, provide your recommended answer.
Ask the questions one at a time.
If a question can be answered by exploring the codebase, explore the codebase instead.
Use this skill whenever the user is building or scaffolding a new Flutter or Dart mobile app, adding any Flutter feature, screen, page, view model, use case, repository, or API/backend call, setting up Flutter push notifications or deep links, or mapping/refactoring an existing Flutter project toward a layered clean architecture — EVEN WHEN the user never names the architecture, rx_notifier, MVVM, clean architecture, get_it, dio, drift, or go_router. It applies the moment the work touches Flutter app structure: a request like "add a login screen", "fetch data from this endpoint", "create a settings feature", "wire up FCM", "open this screen from a notification", or "clean up my Flutter project" is in scope. This skill is THE way to produce Flutter code in this codebase shape: rx_notifier Controller view models, AbstractUseCase with Either<AppError,T> error handling, I*Repository interface/impl pairs, a tiered dio gateway, get_it dependency injection, drift TTL cache, and go_router navigation. Prefer it over a
Knowledge graph semântico da codebase via MCP. Mapeia arquitetura, analisa blast-radius de mudanças e faz code review com até 8x menos tokens. Usa Tree-sitter (24 linguagens) + SQLite.
Mantém uma knowledge base persistente no Obsidian seguindo o pattern Karpathy LLM Wiki — ingest de sources, query sobre wiki compilado, lint de saúde. Dispara quando o usuário pede "ingest", "alimenta o wiki", "atualiza wiki", "consulta wiki", "lint wiki", "/wiki-ingest", "/wiki-query", "/wiki-lint", "que sabemos sobre X", "compila X no wiki", "adiciona ao wiki", ou ao salvar uma nova source em References/.
Analisa projetos Stitch e sintetiza design system semântico em arquivos DESIGN.md.
Transforma ideias vagas de UI em prompts polidos e otimizados para Stitch.
Design system e style guide configurável para projetos frontend. Use SEMPRE que criar componentes, páginas ou UI em React/Next/Angular/Vue.
