Why does grill-me do nothing after I install it?

The grill-me SKILL.md is one line that tells the agent to load the grilling skill. If you installed grill-me on its own, install grilling too. The repo's docs call this out, and installing the whole set avoids the problem.

Can the agent start grill-me on its own?

No. Its frontmatter sets disable-model-invocation: true, so in Claude Code it only runs when you type it. The grilling skill underneath is model-invoked, so the agent may reach for that one when you ask it to grill you.

Can grill-me ask one question at a time instead of rounds?

Yes. The repo's docs say to add the line 'When grilling, ask one question at a time.' to your global CLAUDE.md. Rounds are the default because they cut the number of back-and-forth turns.

What is the difference between grill-me and grill-with-docs?

Both run the same grilling interview. grill-me needs no repo and keeps no state. grill-with-docs also loads the domain-modeling skill, reads your code and records terms in GLOSSARY.md and decisions as ADRs while you talk.

Grill me skill: what it does and how to install it

Q: What is the grill me skill?

grill-me is an Agent Skill in Matt Pocock's mattpocock/skills repo. When you type /grill-me, the agent interviews you about a plan, design or idea in rounds of numbered questions, each with a recommended answer, until every decision is settled. It writes no files.

Source: mattpocock/skills · Maintainer: community: Matt Pocock · License: MIT · Checked: 23 October 2026

What is the grill me skill?

The grill me skill makes your coding agent interview you about a plan, design or idea, round after round, until every decision is made and nothing is left assumed. It is for anyone who wants to think a change through before an agent starts building it, and it works on non-code subjects too.

It lives in mattpocock/skills, a collection Matt Pocock describes in the README as "My agent skills that I use every day to do real engineering - not vibe coding." The skill's own description is short: "A relentless interview to sharpen a plan or design."

The README calls /grill-me and its sibling /grill-with-docs "my most popular skills". As of 23 October 2026, its skills.sh page listed about 1.4 million installs.

What does it do?

The grill-me folder is tiny. Its SKILL.md has one line of instructions: call the Skill tool with "grilling". All of the interview logic lives in a second skill, grilling, which grill-me hands off to. So you need both.

What's in the folder:

SKILL.md: frontmatter with disable-model-invocation: true, plus the one-line hand-off to grilling.
agents/openai.yaml: Codex metadata (display name "Grill Me") with allow_implicit_invocation: false, so Codex also waits for you to call it.

The grilling skill it loads runs the session like this:

Design tree. The agent maps your subject as decisions with other decisions hanging off them.
Rounds, not single questions. Each round asks every decision whose prerequisites are already settled (the "frontier"). Questions that depend on an answer still open wait for a later round.
A fixed format. Each question is numbered and titled, followed by the agent's recommended answer, worded so that "yes" accepts the recommendation. You can reply by number.
Facts versus decisions. Facts the agent can find itself (files, tools) go to a sub-agent instead of to you. Decisions are always yours.
A confirmation gate. The session ends when the frontier is empty, and the agent should not act until you confirm you share the same understanding.

The grill-me docs page adds that the skill is stateless: it writes no files, and the only output is a clearer version of your idea.

How do you install the grill me skill?

The README offers a plugin or editable files and says to pick one per agent, because installing both gives you every skill twice.

Claude Code plugin (whole set, updates itself). The README's command installs the full collection from Anthropic's official plugin marketplace:

claude plugin install mattpocock-skills@claude-plugins-official

Claude Code namespaces plugin skills as /plugin-name:skill-name, so look for it as /mattpocock-skills:grill-me in the slash menu.

Codex plugin. From the README:

codex plugin marketplace add mattpocock/skills
codex plugin add mattpocock-skills@mattpocock

Just these skills, as editable files. The skills CLI accepts --skill more than once, so install grill-me with the grilling skill it depends on:

npx skills add mattpocock/skills --skill grill-me --skill grilling

Add -g to install for every project, or -a claude-code to skip the agent prompt. For Claude Code the folders land in .claude/skills/grill-me/ in your project, or ~/.claude/skills/grill-me/ with -g, which are the standard Claude Code skill locations. Codex and Cursor read the same SKILL.md format; see Codex skills and Cursor skills.

claude.ai. Custom skills upload as a ZIP of the skill folder under Customize, then Skills (Claude Help Center). Since grill-me only points at grilling, upload both folders.

How do you use it?

Type /grill-me followed by your idea, or type it alone and describe the idea when asked. The docs recommend a fresh conversation, not one where an agent has already written a plan, and leaving plan mode off, because plan mode pushes the agent to produce a plan instead of asking questions.

Example prompts (suggestions, not quotes from the repo):

/grill-me I want to add team accounts to our app, with shared billing.
/grill-me Should I move our marketing site from a CMS to Markdown files in the repo?
/grill-me I'm planning a talk on testing for a local meetup. Help me decide what it covers.
/grill-me We're thinking of raising prices for existing customers next quarter.

There are no packages, CLIs or API keys to install. The docs do say the model matters more here than for most skills, so use your strongest one for grilling.

When a session ends and the subject turns out to be software, the docs suggest handing the same conversation to the repo's to-spec skill rather than starting fresh, since the context you just built is the value.

Tips and limits

From the repo's docs, not from testing:

You own the scope. The failure mode the docs warn about is answering "agreed" to forty questions and ending up with the agent's plan. Push back, flag drift, and say "I don't know" when that's true.
Count rounds, not questions. The docs call forty-six questions across four rounds an ordinary session. Hundreds of questions usually means the scope was too big; split the work and grill each piece.
Some questions can't be talked through. Layout and "how should this feel" questions need something to react to. The docs suggest building a throwaway version with the repo's prototype skill, then coming back.
Prefer one question at a time? Add When grilling, ask one question at a time. to your global CLAUDE.md, as the grilling docs describe.
It may start building early. The grilling docs say weaker or faster models sometimes skip the confirmation gate. Their fix is a line in your AGENTS.md or CLAUDE.md telling the agent not to implement without permission.
Working in a codebase? Use grill-with-docs instead. It runs the same interview but also loads domain-modeling and records terms in GLOSSARY.md and decisions as ADRs.

The repo is under the MIT License, so you can copy and edit the skills freely as long as you keep the copyright notice.

Save it in your Kipt library

In Kipt, open Skills and paste the skill's GitHub folder link into the import box. Kipt reads the SKILL.md and its files; for a repo with many skills, like this one, it shows a list to pick from, so you can take grill-me and grilling together. You can also drop the folder or a .zip. Your skills live in your own private GitHub repo and install into Claude Code or Codex with one command:

npx skills add https://github.com/<you>/<repo>.git --all -g

Kipt is not affiliated with Matt Pocock or mattpocock/skills.

Similar skills

Domain modeling skill: the other half of grill-with-docs, which keeps a project glossary and ADRs.
TDD skill: Matt Pocock's red-green testing reference for the build step after you've grilled the plan.
Superpowers skills: a different workflow whose brainstorming skill also questions you before any code is written.
Best Claude Code skills: a wider list to browse.