grill-me

grill-me is a skill for Claude Code from murilomn58/Claude-Spec-Driven-Fase-Zero. It costs 223 tokens per session (2,749 once invoked), scanned A, original, MIT.

A requirements interview for a new feature or a change with unclear scope. It challenges assumptions and creates shared understanding before coding starts.

In plain words
What is it for?
Use it to clarify large or uncertain features, explore possible approaches, and produce an agreed brief before writing a specification or implementation.
Why use it?
It reduces the risk that the developer and coding agent mean different things by the request. Finding gaps early is cheaper than discovering them in finished code.

Skill for Claude Code

Written for Claude Code: shipped in a Claude Code plugin. Also seen: mentions CLAUDE.md; names the AskUserQuestion tool.

Part of the adp-fase-zero plugin — 2 skills shipped together

Good fit Use it to clarify large or uncertain features, explore possible approaches, and produce an agreed brief before writing a specification or implementation.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/murilomn58/claude-spec-driven-fase-zero/grill-me
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent
npx skills add murilomn58/Claude-Spec-Driven-Fase-Zero --skill grill-me
Clone the repo
git clone --depth 1 https://github.com/murilomn58/Claude-Spec-Driven-Fase-Zero

Made for: Claude Code.

Or install adp-fase-zero, the plugin that ships this one along with the rest of its 2 skills.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for grill-me

README.md
[![agentmods](https://agentmods.dev/badge/skills/murilomn58/claude-spec-driven-fase-zero/grill-me/github.svg)](https://agentmods.dev/skills/murilomn58/claude-spec-driven-fase-zero/grill-me)
Your own site
<a href="https://agentmods.dev/skills/murilomn58/claude-spec-driven-fase-zero/grill-me"><img src="https://agentmods.dev/badge/skills/murilomn58/claude-spec-driven-fase-zero/grill-me/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for grill-me

Your own site · 80×15
<a href="https://agentmods.dev/skills/murilomn58/claude-spec-driven-fase-zero/grill-me"><img src="https://agentmods.dev/badge/skills/murilomn58/claude-spec-driven-fase-zero/grill-me.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 223 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,749 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00223 $0.02749
Opus 5 $0.00112 $0.01375
Sonnet 5 $0.00045 $0.00550
Haiku 4.5 $0.00022 $0.00275

Measured 11d ago against content hash f1033680b208, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-11, from the pricing page.

Security

Grade A, and why

grill-me scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/grill-me/SKILL.md · 90 lines

How it starts

The opening of the file, as written. The whole thing — 90 lines — stays where its author put it; the contents beside it link to each section on GitHub.

grill-me

O maior modo de falha no desenvolvimento com agente e o desalinhamento: voce acha que pediu uma coisa, o agente entendeu outra, e isso so aparece no codigo pronto. O grill-me mata o desalinhamento na origem: uma entrevista implacavel que forca entendimento compartilhado ANTES de qualquer linha de codigo. E a Fase 0 do ADP, disparada pelo /adp-spec ou avulsa.

Reescrita da ideia do grill-me do Matt Pocock (mattpocock/skills), adaptada ao ADP com abertura criativa e integracao ao fluxo do protocolo.

SUB-SKILL RECOMENDADA (opcional — sem ela, siga o processo descrito nesta própria skill): superpowers:brainstorming e o motor de descoberta (uma pergunta por vez, propor abordagens, HARD-GATE: nada de implementar sem design aprovado). Esta skill adiciona por cima a POSTURA de grill, a abertura criativa e a saida estruturada do ADP.

Quando usar e quando NAO usar

  • Usar: feature nova, escopo grande, requisito ainda vago, mudanca que cruza varias areas.
  • Nao usar: mudanca pequena, direta e ja compreendida. Nesse caso, va de plan mode direto (regra spec-driven vs plan mode da aula). Nao force spec no trivial.

Postura: grill, nao brainstorm gentil

A diferenca do brainstorming puro esta na postura. Aqui:

  • Uma pergunta por vez. Nunca despeje um questionario. Cada resposta abre a proxima pergunta.
  • Nao aceite resposta vaga. "Sei la", "tipo um dashboard", "o de sempre" nao passam: puxe o "por que" ate o requisito ficar concreto e testavel.
  • Desafie cada suposicao, inclusive as do próprio usuário. Se um pedido colide com o que ja existe no codebase ou com uma restricao real, aponte antes de seguir.
  • Prefira multipla escolha quando der: e mais facil de responder e expoe as opcoes.
  • Pare quando afiado, nao antes. Escopo, criterios de sucesso e fronteira (o que fica de fora) precisam estar claros.

O loop do grill

  1. Dupla exploracao em paralelo (2 subagentes de uma vez, ANTES de perguntar). Despache no MESMO turno:
    • Terreno interno: UM subagente pra mapear o codebase do ALVO (caminho absoluto): feature-dev:code-explorer quando o projeto tem estrutura de app; fallback para um Explore com Glob/Read quando nao se aplica (ex: site estatico vanilla). NUNCA explore inline com ls/find no thread principal, e nunca repita um comando que falhou. No mesmo passo, cheque reuso interno: memoria da sessao, CLAUDE.md dos projetos e Glob bounded em C:\dev\* — muita "feature nova" ja existe como modulo pronto (ex: um modulo de pagamento seu, pronto em outro repo). Cite arquivos.
    • Radar largo (estado da arte, semeado so pelo prompt inicial): UM subagente com a web-buscador no playbook "Radar estado-da-arte", passada larga: como o mundo resolve isso HOJE — abordagens dominantes e emergentes, stack de referencia (versoes atuais, o que a comunidade adotou/abandonou), templates e repos exemplares, concorrentes lancados, dores relatadas. Cobre o radar tecnico E o de comunidades (Reddit, dev.to, Product Hunt, TabNews, HN, GitHub). Voce entra na entrevista JA sabendo o terreno interno e o estado da arte — as perguntas nascem melhores.
  2. Grelhar (informado pelo radar). Perguntas dirigidas, uma por vez: problema real, quem usa, criterios de sucesso, casos de borda, restricoes (stack, prazo, dado sensivel, LGPD). Use os achados como MUNICAO de pergunta, nao como palestra: "a maioria resolve isso com X; teu caso tem motivo pra diferir?", "o produto Y ja faz isso; o que te diferencia?", "a comunidade abandonou Z por causa de W; ainda quer seguir por ai?". Registre as respostas.
  3. Abrir opcoes (a criatividade entra aqui). Antes de fixar o caminho, monte 2-4 abordagens (stacks, libs, MCPs, plugins, padroes atuais) com trade-offs, cruzando o radar largo com o que a sua equipe ja tem (skills, MCPs). Se as respostas do grill mudaram a direcao (nicho diferente, restricao nova), dispare uma segunda passada FINA da web-buscador, agora mirada na direcao escolhida — curta e cirurgica, nada de repetir a larga. Nao fixe o caminho sem ter aberto as opcoes.
  4. Fundir (achado vira decisao, nunca trivia). Cruze CADA achado relevante do radar com o escopo pedido no prompt E com as respostas do brainstorm, e de a ele um unico destino (tabela na secao Fusao). Achado sem destino = corta. E aqui que o estado da arte vira escopo, e nao curiosidade.
  5. Afiar o dominio. Conforme as decisoes cristalizam, registre os termos-chave (glossario leve) e as decisoes importantes com o porque (ADR embrionaria). Isso vira materia-prima do PRD e evita re-discussao. (inspirado no domain-modeling do Matt)
  6. Convergir. Proponha a abordagem recomendada com a justificativa + as alternativas como plano B.

Read the full file on GitHub · 90 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 11d ago First seen · 90 lines · 223 tokens per session scan A f1033680b208

Subscribe to this mod's changes

grill-me is a skill published in the GitHub repository murilomn58/Claude-Spec-Driven-Fase-Zero (1 stars, last pushed 2mo ago), licensed MIT. It adds 223 tokens to every session and 2,749 once invoked, about $0.0011 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other skills, from other repositories

justone

Runtime contract for the Just One Prompt Challenge. Invoked exclusively via the /justone and /justone-publish slash commands. Do NOT auto-trigger from general conversation — only act on this skill when one of those commands has been issued in the current turn.

GodsB1025/justone · 55 tokens

memstack-security-owasp-top10

Use this skill when the user says 'OWASP audit', 'OWASP top 10', 'security audit', 'vulnerability assessment', 'full security check', or needs a comprehensive web application security review against OWASP Top 10 categories. Do NOT use for dependency audits or secret scanning alone.

cwinvestments/memstack · 69 tokens

memstack-business-gdpr

Use this skill when the user says 'GDPR', 'data protection', 'privacy compliance', 'DPA', 'DSAR', 'data subject request', 'cookie consent', 'privacy audit', 'CCPA', or asks 'do I need GDPR for this repo'. Scans the repository to detect what personal data is collected, classifies sensitivity, determines whether GDPR…

cwinvestments/memstack · 121 tokens

memstack-business-scope-of-work

Use this skill when the user says 'scope of work', 'SOW', 'define scope', 'project scope', 'write SOW', 'scope document', or is defining project boundaries, deliverables, and acceptance criteria for a formal engagement. Do NOT use for proposals, contracts, or invoicing.

cwinvestments/memstack · 70 tokens

memstack-content-youtube-script

Use this skill when the user says 'YouTube script', 'video script', 'write script for YouTube', 'YouTube video outline', or is creating scripted content for a YouTube video with hooks, chapters, and CTAs. Do NOT use for TikTok/Reels short-form scripts or webinar presentations.

cwinvestments/memstack · 70 tokens

memstack-seo-local-seo

Use this skill when the user says 'local SEO', 'Google Business Profile', 'local search', 'NAP consistency', 'local listings', 'Google Maps', 'local pack', or is optimizing a business for local search results and map visibility. Do NOT use for general SEO audits or national keyword research.

cwinvestments/memstack · 69 tokens