How do I install the Wheel Grilling skill?

Run npx skillmds@latest add kalpakprod/wheel-grilling in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Wheel Grilling safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Wheel Grilling?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Wheel Grilling free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Wheel Grilling?

kalpakprod (@kalpakprod) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Wheel Grilling by kalpakprod — Agent Skill

Name: Wheel Grilling
Author: kalpakprod

Wheel Grilling

Internal Wheel questioning stage that reduces live-researched Families and Candidates. Use only from wheel after its quick pass or after one user answer; emit exactly one legal next question or declare the Candidate set stable.

kalpakprod cf64913 2.5 KB Updated Sep 17, 2026

File contents

wheel-grilling

Turn the researcher's Families, Candidates, Evidence, and decision axes into one user decision at a time. wheel-grilling is the sole owner of question construction and emission. Facts are the researcher's job. User preferences decide among fact-backed options.

Legal question

A question is legal only when at least one possible answer changes the surviving Families, Candidates, possible Core, or integration method. Never ask about an implementation detail already supplied by every remaining Candidate.

Emit exactly one question per user turn. It includes:

the choice in plain language;
answer options;
Families and Candidates affected by every option;
a recommendation grounded in a verified fact;
the Candidate-count consequence when known.

Question: Managed service or self-hosted Core?
Managed service: affects Family A; keeps Candidate 1 and Candidate 2.
Self-hosted: affects Family B; keeps Candidate 3 and Candidate 4.
Recommendation: self-hosted, because the project instruction requires local data control.

Do not present a recommendation as a user preference. Do not ask a second question while waiting for an answer.

Flow

After the Quick pass, derive and emit the next legal question from the supplied three to six Families, Candidates, Evidence, and decision axes.
After each answer, update requirements and request bounded research only for the affected branch.
Recompute the remaining Candidates, Candidate roles, and integration methods before proposing another question.
Continue only while a legal answer can change the final set.
Declare grilling stable when no remaining answer changes the final Candidate set, or when the user explicitly delegates the choice to Wheel.

If no Candidate remains, keep grilling and hand every new constraint, synonym, adjacent niche, or English term to wheel-research as a new search lead. Zero Candidates never authorizes a new implementation.

Handoff

Return one of:

question: one legal question with affected Families and Candidates;
stable: remaining Candidates, answered constraints, eliminated Candidates and reasons, and any unresolved verification.

Wheel starts deep research only from stable.

---name: wheel-grillingdescription: Internal Wheel questioning stage that reduces live-researched Families and Candidates. Use only from wheel after its quick pass or after one user answer; emit exactly one legal next question or declare the Candidate set stable.---# wheel-grillingTurn the researcher's Families, Candidates, Evidence, and decision axes into one user decision at a time. `wheel-grilling` is the sole owner of question construction and emission. Facts are the researcher's job. User preferences decide among fact-backed options.## Legal questionA question is legal only when at least one possible answer changes the surviving Families, Candidates, possible Core, or integration method. Never ask about an implementation detail already supplied by every remaining Candidate.Emit exactly one question per user turn. It includes:- the choice in plain language;- answer options;- Families and Candidates affected by every option;- a recommendation grounded in a verified fact;- the Candidate-count consequence when known.```textQuestion: Managed service or self-hosted Core?Managed service: affects Family A; keeps Candidate 1 and Candidate 2.Self-hosted: affects Family B; keeps Candidate 3 and Candidate 4.Recommendation: self-hosted, because the project instruction requires local data control.```Do not present a recommendation as a user preference. Do not ask a second question while waiting for an answer.## Flow1. After the Quick pass, derive and emit the next legal question from the supplied three to six Families, Candidates, Evidence, and decision axes.2. After each answer, update requirements and request bounded research only for the affected branch.3. Recompute the remaining Candidates, Candidate roles, and integration methods before proposing another question.4. Continue only while a legal answer can change the final set.5. Declare grilling stable when no remaining answer changes the final Candidate set, or when the user explicitly delegates the choice to Wheel.If no Candidate remains, keep grilling and hand every new constraint, synonym, adjacent niche, or English term to `wheel-research` as a new search lead. Zero Candidates never authorizes a new implementation.## HandoffReturn one of:- `question`: one legal question with affected Families and Candidates;- `stable`: remaining Candidates, answered constraints, eliminated Candidates and reasons, and any unresolved verification.Wheel starts deep research only from `stable`.

Frequently asked questions