Guide · Agent skill
How to use the Grilling Plan Stress-Tester agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The grilling skill changes how an agent responds when you ask it to stress-test a plan, decision, or idea. Instead of giving quick feedback, the agent interviews you in numbered rounds until you share an understanding of the plan. This guide is for people who have not used agent skills before, and it covers what the skill does, how to install it, how to trigger it, and where it fits best.
Quick start
- 1Download the ZIP. Get the grilling skill ZIP from this directory's page.
- 2Unzip into a skills folder. Extract it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder.
- 3Or upload in the Claude apps. Go to Settings → Capabilities → Skills and upload the ZIP.
- 4Describe the plan. Tell the agent what plan, decision, or idea you want stress-tested.
- 5Answer each round. Reply to the numbered questions, accepting or correcting the recommended answers.
- 6Confirm shared understanding. When the questions run out, confirm the shared understanding before the agent acts.
What is the grilling skill and what does it make the agent do?
Agent skills are instruction packages that tell an agent how to handle a specific kind of task. The grilling skill is licensed under MIT and is described as a way to grill you relentlessly about a plan, decision, or idea.
Once installed, the agent maps your decisions as a design tree, where each decision branches into the decisions that depend on it. It asks the questions that can be answered right now, numbers them, and gives a recommended answer to each. It then waits for your replies before going deeper, and it keeps going until no open questions remain.
- It groups questions into rounds and asks only the ones whose prerequisites are settled.
- Each question is worded so that answering "yes" accepts the agent's recommended answer.
- If it needs a fact from your environment, it looks the fact up itself rather than asking you.
- It will not act on the plan until you confirm you have reached a shared understanding.
How do you install the grilling skill?
You can install the skill in two ways. The first is to download the ZIP file from the skill's page on this directory and unzip it into your Claude skills folder. Use ~/.claude/skills/ for a personal install available across your projects, or a project's .claude/skills/ folder to limit it to one project.
The second is to upload the skill through the Claude apps. Open Settings, go to Capabilities, then Skills, and upload the ZIP there.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into the project's .claude/skills/ folder
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
- The package contains agents/openai.yaml and SKILL.md
How do you trigger the grilling skill with a natural request?
You do not need a special command. The skill is meant to activate when you ask the agent to stress-test your thinking, or when you use a "grill" trigger phrase. Describe what you want to be challenged on and say so plainly.
The skill's description lists the trigger as the user wanting to stress-test their thinking or using any grill trigger phrases. The source does not list specific phrases, so the wording above is the safest guide.
- "Grill me on this plan before I commit to it."
- "Stress-test my decision to move the billing service to a new database."
- "I have an idea for a note-taking app. Grill me on it."
What happens during a grilling session?
The first reply is a numbered round of questions, each with a recommended answer marked by an arrow. Reply to the questions you agree with, correct the ones you do not, and the agent recomputes what it can ask next. Answers can unlock questions that were waiting on them, so later rounds often go deeper into details.
The session ends when the open questions run out. Only then should you expect the agent to move toward acting on the plan, and only after you confirm the shared understanding.
- Expect several rounds before any action is taken.
- You can reply with the answers you have and leave the rest for later rounds.
- The agent is designed to put decisions to you rather than make them for you.
What does the skill need and what are its safety notes?
The source lists no requirements. It does not run scripts and does not use the network. Because of that, the skill is a set of instructions rather than code that runs on your machine.
The one behaviour to plan for is the process itself. The agent asks many questions before doing anything, so it is slow by design.
- Requires: nothing listed
- Runs scripts: no
- Uses network: no
- Sub-agents may be used to look up environment facts, as the skill instructs
What people use it for
- Pressure-testing a product idea before building it
- Surfacing assumptions in a feature or architecture plan
- Working through a major business or project decision step by step
- Checking a plan for edge cases you tend to skip
Frequently asked questions
Is the grilling agent skill free to use?
The skill is published under the MIT licence, according to this listing. Check the licence file in the source repository for the full terms.
What do I need to run the grilling skill?
The source lists no requirements. It does not run scripts and does not use the network, so it is instruction-only.
Where do I install an agent skill?
Unzip the ZIP into ~/.claude/skills/ for personal use or a project's .claude/skills/ folder. In the Claude apps, upload it under Settings → Capabilities → Skills.
Why does the agent keep asking me questions instead of answering?
That is the intended behaviour. The skill asks questions in rounds and waits for your replies, and it will not act on the plan until you confirm shared understanding.
Can I skip the questions and just get a quick answer?
Not with this skill. It is deliberately slow and repetitive, so for quick answers or simple tasks you should use an agent without it or ask a plain question.
How is this different from asking an agent to review my plan?
A normal review usually gives feedback in one reply. This skill interviews you in numbered rounds, keeps a design tree of decisions, and keeps going until no open questions remain.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use the grilling skill if you want your thinking challenged before a project, product, or major decision, or if you tend to miss edge cases. Pick something else if you want fast answers, want the agent to make decisions and proceed, or find repeated questioning frustrating.
Download the skill