grill-plan
Skill0 starsChanged 37 days agoMITView on GitHub
Interrogate a plan, design or decision until nothing is silently assumed: map the decisions as a tree, ask the whole settled frontier in one numbered round with a recommended answer per question, look facts up instead of asking for them, put only decisions to the decider, and record each settled decision in the plan or as a decision record. Triggers on: "grill me", "grill this", "stress-test", "poke holes", "interrogate the plan", "what am I missing", "challenge this", "before I commit to this".
5 other projects use a version of this.
Adopted by
Most recently taken up by
- microsoft/agentic-journeystheir own version, first seen Sep 30, 2026
- ai-agent-lead/skillstheir own version, first seen Sep 26, 2026
- guobinmengxiang-rgb/skillstheir own version, first seen Aug 16, 2026
- ltatarev/skillstheir own version, first seen Aug 9, 2026
- Jinx-1120/skillstheir own version, first seen Jul 29, 2026
Other versions5 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 5
- microsoft/agentic-journeys★ 5same name
- ai-agent-lead/skills★ 3same name
- guobinmengxiang-rgb/skills★ 1same name
- Jinx-1120/skills★ 1same name
- ltatarev/skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository11 other files it gives its agents.
- AGENTS.mdAGENTS.md
- cut-ticketsSkill
- decision-recordsSkill
- deep-modulesSkill
- evidenceSkill
- plain-voiceSkill
- plan-projectSkill
- review-changeSkill
- setup-conventionsSkill
- test-firstSkill
- thin-slicesSkill
The file
Files
Grill a plan
Interview until there is a shared understanding and nothing is left assumed. The decisions form a tree: every decision branches into the decisions that hang off it. Work the tree in rounds.
The rules
| Id | Rule | In practice |
| GRILL-1 | Rounds over the frontier | The frontier is every decision whose prerequisites are settled. Ask the whole frontier in one round: number each question, give a recommended answer, word the question so "yes" accepts the recommendation. A question that depends on another one still open waits for the next round. |
| GRILL-2 | Facts are the agent's job | Anything the filesystem, the repository, a document or a tool can answer is looked up, never asked. A running lookup is an unsettled prerequisite: ask the rest of the frontier now, hold only the questions downstream of it. |
| GRILL-3 | The session ends when the frontier is empty | Every branch visited, nothing silently assumed. Act on nothing until the decider confirms the shared understanding. Each settled decision goes into the plan's Decisions table (PLAN-7) or, when it is hard to reverse and surprising, a decision record (DECIDE-1); when there is no plan, the conversation summary. |
A round
Q1. <title>: <the question, with the options when there are several>
Recommended: <answer>. Yes accepts it.
Q2. <title>: <question>
Recommended: <answer>.Plain text, no emoji, no decoration. After the answers, recompute the frontier: settled decisions unblock the questions that depended on them. Ask the next round.
What a good question does
- It separates a decision (the decider's) from a fact (yours to find).
- It names the consequence of each option in one line, tagged Verified, Measured or Inferred (EVID-1).
- It is specific enough that "yes" is an answer.
Companion skills
| Need | Skill |
| Where the settled decisions go | plan-project, decision-records |
| The vocabulary for interface and seam questions | deep-modules |
| Calibrating the consequences you state | evidence |
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

