grill-plan
Skill5 starsChanged 46 days agoMITView on GitHub
Interview the user about an issue or plan before any code is written, one question at a time, until the decisions and the test list are clear. Turns a plan into a Decisions table and a behavior-to-test list for red/green TDD. USE FOR: grill me, grill the plan, stress-test a plan, plan interview, clarify an issue before coding, find ambiguous requirements, decide defaults before implementation. DO NOT USE FOR: writing code or tests (use the tdd-builder agent), reviewing finished code (use /review), or creating new journeys (use journey-template).
4 other projects use a version of this.
Adopted by
Most recently taken up by
- 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 versions4 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 4
- 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
- journeys/smart-todo/setup/copilot-instructions.mdCopilot instructions
- container-apps-deploymentSkill
- data-access-abstractionSkill
- grafana-azureSkill
- journey-runnerSkill
- journey-templateSkill
- journey-test-harnessSkill
- n8n-azureSkill
- superset-azureSkill
- flex-functions-sql-foundrySkill
The file
We split this skill into the rules it gives an agent
9 rules. 0 are common, 7 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 4 edited versions we read keep these.
7
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 15
Read first: Read the issue and every plan document and section it links to.
- Very specific: line 16
Mandatory questions: If a linked plan has a Decision Points section, ask about every item in it.
- Distinctive: line 18
One at a time: Ask each question with the ask-user tool.
- Distinctive: line 19
Summarize: When the questions are done, print:
- Very specific: line 23
Record it: When the user asks, post the summary as a comment on the issue with gh issue comment <number> --body-file <file>.
- Distinctive: line 31
Do not write application code or tests.
- Distinctive: line 32
Do not add requirements the plan and issue do not support.
- Distinctive: line 33
If the issue and the plan conflict, ask which one wins.
- Distinctive: line 34
Keep each question under 60 words.
“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Grill the Plan
Find the decisions a plan leaves open before an agent fills them in silently. The output is a short list of decisions and the tests that will prove them.
Process
- Read first. Read the issue and every plan document and section it links to. Do not ask questions the plan already answers.
- Mandatory questions. If a linked plan has a
Decision Pointssection, ask about every item in it. Each item lists a default. Offer that default as the first, recommended option. - Your own questions. Then ask up to three more questions about gaps you found yourself. Good targets are boundary values, failure paths, what happens to related data, and how each acceptance criterion will be tested. Skip questions whose answer would not change the code or the tests.
- One at a time. Ask each question with the ask-user tool. Include two to four options, put the recommended default first, and add one sentence on why the choice matters. If the user says "use the defaults", accept every remaining default and stop asking.
- Summarize. When the questions are done, print:
- A
Decisionstable with the columns Question, Decision, and Default or changed. - A
Test listwith one row per behavior (Behavior, Plan section, Proposed test name). The tdd-builder agent writes these tests in the red phase. - Anything still unresolved.
- A
- Record it. When the user asks, post the summary as a comment on the issue with
gh issue comment <number> --body-file <file>. Write the body to a temporary file first so the command works in PowerShell, Command Prompt, bash, and zsh. Delete the temporary file afterward.
Non-interactive sessions
If the ask-user tool isn't available, as in copilot -p or a cloud agent session, don't pretend an interview happened. Apply the recommended default for every question, mark each plan Decision Point row in the Decisions table as Default (not asked) and each question you raised yourself as Agent recommendation (not asked), and start the summary with one sentence saying that the interview ran non-interactively.
Rules
- Do not write application code or tests.
- Do not add requirements the plan and issue do not support. If you think one is missing, ask about it as a question.
- If the issue and the plan conflict, ask which one wins.
- Keep each question under 60 words.
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.