grill-me
Skill3 starsChanged 5 months agoMITView on GitHub
Use PROACTIVELY whenever the user is working something OUT OF THEIR HEAD rather than asking a quick question — mulling or scoping a plan/design/decision, describing how they do something, or floating a half-formed idea to capture. Natural signals: "I've been thinking about…", "I have an idea for…", "I want to build/plan/figure out…", "let me think this through", "help me scope this", "here's how I do X". Also the explicit "grill me", "interrogate me", "stress-test this", "/grill-me". One-question-at-a-time extraction into a durable doc. SAFETY: unless they explicitly asked to be grilled, OFFER first and wait for yes — never silently launch a long interview on a quick question.
193 other projects use a version of this, 10 of them edited.
Adopted by
Most recently taken up by
- bamlab/flashlighttheir own version, first seen Nov 16, 2026
- philoserf/claude-code-configtheir own version, first seen Nov 15, 2026
- niki914/zafirotheir own version, first seen Nov 15, 2026
- quickchatai/quickreviewtheir own version, first seen Nov 15, 2026
- HiteSit/real-scientist-skillstheir own version, first seen Nov 15, 2026
- alexj11324/Mikkatheir own version, first seen Nov 15, 2026
Used with
Projects with it also use brief-to-tasks, writing-beats and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
julianoczkowskiSkillbrief-to-tasks
Break a design brief into an ordered checklist of independently buildable tasks using vertical slices. Saves as a markdown checklist. Use when user wants to break down work, create tasks from a brief, plan implementation order, or mentions "tasks" or "breakdown".
Also in 7 projects that use this
LesaniSkillwriting-beats
Shape an article as a journey of beats, choose-your-own-adventure style. The user picks a starting beat from the raw material, you write only that beat, then offer options for where to pivot next, beat by beat, until the article reaches a natural end. Use when the user has raw material and wants to assemble it as a narrative rather than an argument.
Also in 74 projects that use this
simo97Skillmigrate-to-shoehorn
Migrate test files from `as` type assertions to @total-typescript/shoehorn. Use when user mentions shoehorn, wants to replace `as` in tests, or needs partial test data.
Also in 78 projects that use this
simo97Skillrequest-refactor-plan
Use this skill when user wants to create a refactor plan.
Also in 45 projects that use this
LesaniSkillwriting-shape
Take a markdown file of raw material and shape it into an article through a conversational session — drafting candidate openings, growing the piece paragraph by paragraph, arguing about format (lists, tables, callouts, quotes) at each step. Use when the user has a pile of notes, fragments, or a rough draft and wants help turning it into something publishable.
Also in 73 projects that use this
RoentekSkillloop-me
Grill me about specs for the workflows I want to build, within this workspace.
Also in 64 projects that use this
- +1 more from simo97
- +1 more from Lesani
Other versions193 projects use a version of this.
Word for word · 0
None found.
Edited · 10
- openstatusHQ/openstatus★ 9.3k10% same
- sanity-io/sanity★ 6.4k10% same
- camunda/camunda★ 4.4k10% same
- openstatusHQ/data-table-filters★ 2.3k10% same
- nateherkai/AIS-OS★ 1.4k16% same
- jewbetcha/openflight★ 1.2k10% same
- truffle-ai/dexto★ 66110% same
- brianlovin/claude-config★ 37610% same
- kubev2v/forklift★ 21310% same
- mart337i/odoo-skills★ 4311% same
Wrote their own version · 183
- mattpocock/skills★ 325ksame name
- The-Vibe-Company/quivr★ 40ksame name
- QuivrHQ/quivr★ 40ksame name
- langfuse/langfuse★ 37ksame name
- nrwl/nx★ 29ksame name
- windmill-labs/windmill★ 19ksame name
- GreyDGL/PentestGPT★ 16ksame name
- humanlayer/humanlayer★ 13ksame name
- jaredpalmer/kev★ 9.8ksame name
- webiny/webiny-js★ 8.1ksame name
173 more not listed
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 repository15 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- agent-builderSkill
- agents-team-builderSkill
- aios-auditSkill
- hooks-builderSkill
- level-upSkill
- multi-brainSkill
- onboardSkill
- plugin-builderSkill
- printing-pressSkill
- routines-builderSkill
- session-handoffSkill
- skill-builderSkill
- workflow-builderSkill
The file
We split this skill into the rules it gives an agent
11 rules. 0 are common, 9 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 179 edited versions we read keep these.
9
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 31
Don't collide with neighbors: exploring brand-new design space → brainstorming.
- Distinctive: line 47
After EVERY answer, BEFORE the next question:
- Distinctive: line 49
Update earlier entries: if a later answer changes them.
- Distinctive: line 50
Only then: ask the next question. Never batch multiple answers into one write.
- Distinctive: line 56
Explore instead of asking: If a question can be answered by reading a file or the wiki, do that instead.
- Distinctive: line 57
Flag and move on: When only a stakeholder can answer, capture a flag with the owner and continue.
- Distinctive: line 82
Final reconciliation pass: read the whole capture for contradictions or gaps.
- Very specific: line 85
If deferred, the capture stays in knowledge/grills/ and you log a follow-up in pending.md.
- Distinctive: line 90
Never skip a checkpoint. The file is the source of truth, not your context window.
- Distinctive: line 91
Don't invent answers — flag what only a stakeholder knows.
- Very specific: line 92
No parallel knowledge store: captures in knowledge/grills/, polished knowledge graduates into context/ / references/.
“Common” is measured: kept by at least 6 in 10 of the 179 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
What this skill does
Relentlessly interview the user about every aspect of a topic until you reach shared understanding. Walk down each branch of the decision tree, resolving dependencies one by one. The real goal is to extract what's in their head into a durable, organized markdown file so nothing is lost as context fills up — and then graduate that capture into the wiki so the knowledge compounds.
This is the general-purpose extraction engine. The builders (/skill-builder, /agent-builder, …) each run a Discovery Interview tied to building one artifact. This skill grills the user about anything — a business process, a plan, a design, a decision — for its own sake, and files the result.
Pattern credit: original "grill me" prompt by Matt Pocock; checkpointing fork by Nate Herk. This is the AIOS adaptation: captures live in knowledge/grills/ and graduate into context/ or references/ per references/wiki-protocol.md.
When to fire — and offer vs. just go
Watch for the situation, not the phrase: the user is mulling a plan out loud, describing "how I do X," floating a half-formed idea, or scoping something before a build. That's a grill opportunity. But the cost of getting it wrong is asymmetric — silently launching a 30-question interview when they wanted a quick answer is worse than a missed trigger. So:
- They explicitly asked ("grill me", "/grill-me", "stress-test this") → just go.
- You inferred the opportunity → OFFER first, one line, then wait: "Sounds like there's a lot in your head here — want me to grill you on it and capture it into the wiki? (one question at a time)" Only start on a yes.
- It's a genuine quick question → do not fire. No offer.
Don't collide with neighbors: exploring brand-new design space → brainstorming; building one specific artifact → that builder's own discovery interview. Grill-me extracts knowledge and decisions the user already holds.
The capture file is the whole point
Long interviews fill up context. If you hold answers only in your head, you will eventually misremember, conflate, or drop something. So you checkpoint to disk after every single answer. The file, not your context, is the source of truth.
Setup — BEFORE the first question
- Pick the topic + slug. From
$ARGUMENTSif given, else ask one line: "What are we grilling on?" - Get today's date:
date +%F. - Create the capture file at
knowledge/grills/{YYYY-MM-DD}-{topic-slug}.md(create the folder if needed). One predictable home, regardless of topic. - Seed it immediately with the template below.
- Tell the user where you're saving, in one line. Then ask Q1.
The checkpoint rule (non-negotiable)
After EVERY answer, BEFORE the next question:
- Append a structured entry: the question topic, the key facts and decisions (in their words where the wording matters), and any flags (things they couldn't answer + who can).
- Update earlier entries if a later answer changes them. Keep the running Summary current.
- Only then ask the next question. Never batch multiple answers into one write.
Interview method
- One question at a time. For each, provide your recommended answer (your best inference from their context — CLAUDE.md, the wiki, the repo) so they can confirm, correct, or redirect.
- Resolve dependencies in order: settle the upstream decision before the ones that depend on it.
- Explore instead of asking. If a question can be answered by reading a file or the wiki, do that instead.
- Flag and move on. When only a stakeholder can answer, capture a flag with the owner and continue.
- Keep going until they say done or every branch is covered. Near the end: "Anything we haven't touched?"
- Match
references/voice.md— this is a conversation, not an interrogation transcript.
Capture file template
# {Topic}: Grill / Discovery Notes
Date: {date} · Goal: {one line}
## Summary / key decisions
(running synthesis, updated as you go)
## Q&A log
### Q1 — {topic}
- Asked: {question}
- Captured: {facts, decisions}
- Flags: {open item -> owner}
## Open flags (pending input)
- {item} -> {who can answer}At the end — reconcile, then graduate
- Final reconciliation pass: read the whole capture for contradictions or gaps; make the Summary stand on its own.
- Recap in chat: what's captured, what's flagged (with owners), suggested next step.
- Offer to graduate into the wiki, proposing the destination per
references/wiki-protocol.md: personal/work facts → acontext/file; a decision-with-a-why → append todecisions/log.md; fast-changing personal facts or preferences →memory/(if present); evergreen reusable knowledge → areferences/page (then updateindex.md, append tolog.md, and add a row toknowledge/README.md's Source registry: date ·knowledge/grills/<file>· grill capture · graduated to<page>); a build plan → the relevant project spot. Heads-up:references/is committed and public — unlike the gitignoredknowledge/grills/capture and the privacy-gatedcontext/— so only graduate non-sensitive evergreen knowledge there; anything tied to an employer or a specific person goes tocontext/. Let the user confirm or defer — don't auto-write into the wiki without a yes. - If deferred, the capture stays in
knowledge/grills/and you log a follow-up inpending.md.
Guardrails
- One question at a time. Always.
- Never skip a checkpoint. The file is the source of truth, not your context window.
- Don't invent answers — flag what only a stakeholder knows.
- No parallel knowledge store: captures in
knowledge/grills/, polished knowledge graduates intocontext//references/. - This skill stays interactive in the live session — do not fork it to an isolated agent.
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.

