grill
Skill4 starsChanged 12 days agoMITView on GitHub
This skill should be used when the user asks to "grill", "interrogate", "stress-test a spec/plan", "adversarially review a proposal", or invokes /craft:grill. Convergent counterpart to brainstorm — interrogates a spec/plan/topic one question at a time to surface gaps, contradictions, and unresolved dependencies before implementation, then captures a durable GRILL-*.md decision ledger that feeds /craft:plan → /do. For GENERATING options (divergent ideation), use the brainstorm skill instead.
87 other projects use a version of this.
Adopted by
Most recently taken up by
- Yesterday-AI/ytstacktheir own version, first seen Nov 8, 2026
- therealtinhtute/mono-harnesstheir own version, first seen Nov 8, 2026
- gaoyu06/harnesstheir own version, first seen Nov 8, 2026
- afawfawffaf/my-skillstheir own version, first seen Nov 8, 2026
- BSSE23047/nauraiz-devtheir own version, first seen Nov 8, 2026
- Mahesh-Bishnoi/skillstheir own version, first seen Nov 8, 2026
Used with
Projects with it also use write-skill, root-cause 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.
NoaiRoxSkillwrite-skill
Use when creating a new skill, editing an existing skill, or verifying a skill works before deployment.
Also in 5 projects that use this
Ingramradical235Skillroot-cause
Structured root cause analysis using WHERE (MECE decomposition) then WHY (causal structure diagrams) pipeline. Includes coin-flip detection, logical-leap detection, and vicious cycle intervention. Use when buffer enters YELLOW/RED, when metrics stall, or when Actions execute but don't produce results.
Also in 4 projects that use this
Neko-Catpital-LabsSkillprove-it
Shared evidence rule for investigating, diagnosing, debugging, proving, or explaining what happened or why for any problem or situation: do not assert something is fixed, working, passing, merged, or running unless you personally observed it this turn and can show what you saw; do not assert it was caused by X until a repro and controlled isolation prove that cause.
Also in 4 projects that use this
sahinSkillexecute
Also in 5 projects that use this
assisjpSkillfile-pr
Prepare and open or update an authorized pull request with a concise problem-first description and verifiable evidence. Use when the user asks to create, file, open, or update a PR. Do not use merely to review a diff or when publication is not authorized.
Also in 3 projects that use this
Arman-KudaibergenovSkillwrite-plan
Этот скилл MUST быть вызван когда есть design.md и нужно разбить доработку на атомарные задачи (2-5 мин) с точными путями файлов и критериями проверки в tasks.md. SHOULD также вызывать как промежуточный шаг между brainstorm и subagent-dev. Do NOT использовать для обсуждения идеи — используй brainstorm; для выполнения задач — используй subagent-dev.
Also in 3 projects that use this
Other versions87 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 87
- kirodotdev/KiroCrew★ 4.8ksame name
- crbnos/carbon★ 2.8ksame name
- JuliusBrussee/cavekit★ 1.1ksame name
- SethGammon/Citadel★ 927same name
- nagi-studio/skills★ 470same name
- KerberosClaw/kc_ai_skills★ 79same name
- alamops/skills★ 53same name
- Kanevry/session-orchestrator★ 53same name
- oakreader/oakreader★ 46same name
- manusco/resonance★ 38same name
77 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 repository42 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- system-architectSkill
- project-detectorSkill
- audit-routerSkill
- command-skill-token-efficiencySkill
- plugin-auditSkill
- sync-featuresSkill
- backend-designerSkill
- devops-helperSkill
- frontend-designerSkill
- git-workflowSkill
- dist-extrasSkill
- distribution-strategistSkill
- homebrew-formula-expertSkill
- homebrew-multi-formulaSkill
- homebrew-setup-wizardSkill
- homebrew-workflow-expertSkill
- architecture-decision-recordsSkill
- changelog-automationSkill
- claude-md-lifecycleSkill
- guard-auditSkill
- hooksSkill
- insights-applySkill
- mode-controllerSkill
- drive-engineSkill
- plan-orchestratorSkill
- repo-triageSkill
- session-stateSkill
- task-analyzerSkill
- workflow-engineSkill
- orchestrator-resilienceSkill
- project-plannerSkill
- preflight-checkSkill
- releaseSkill
- test-generatorSkill
- test-strategistSkill
- adhd-workflowSkill
- brainstorm-insightsSkill
- brainstormSkill
- prompt-refinerSkill
- background-task-managerSkill
The file
We split this skill into the rules it gives an agent
7 rules. 0 are common, 3 are distinctive, and 4 are very specific to this project.
0
Common
Most of the 69 edited versions we read keep these.
3
Distinctive
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 12
Use grill to stress-test a spec/plan (or a bare topic) until every load-bearing branch of the design tree is resolved, then capture the …
- Very specific: line 44
When the argument is a quoted/bare topic, run prompt-refiner on it first (default-on.
- Very specific: line 47
Before asking anything, explore the codebase — read repo evidence (.STATUS, git log, docs/specs/*, commands/, ~/.claude/settings.json when …
- Distinctive: line 68
Unbounded: stop when the load-bearing branches are resolved — target roughly 5 branches, not an exhaustive tree.
- Very specific: line 72
Ask one question per AskUserQuestion call — one-at-a-time fidelity is the point.
- Distinctive: line 77
EVERY option carries a one-line consequence of choosing it.
- Distinctive: line 85
Milestone checkpoints (ADHD-friendly): every 5 resolved branches, pause with an AskUserQuestion: "keep going / wrap up now / show ledger so …
“Common” is measured: kept by at least 6 in 10 of the 69 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 — Interrogate Before Building
Convergent counterpart to brainstorm. Brainstorm is divergent — it GENERATES options and
expands the space. Grill is convergent — it INTERROGATES a position to find gaps,
contradictions, and unresolved dependencies before you implement.
Use grill to stress-test a spec/plan (or a bare topic) until every load-bearing branch of the
design tree is resolved, then capture the locked decisions as a durable ledger that feeds
/craft:plan → /do.
Boundary With Adjacent Skills
| Skill | Role |
| grill (this) | Convergent — interrogates a position, captures a GRILL-*.md decision ledger |
brainstorm | Divergent — generates options, captures a BRAINSTORM/SPEC |
prompt-refiner | Sharpens a bare/quoted topic before grilling (default-on for topics) |
plan-orchestrator / --orch | Turns the locked ledger + spec into an ORCHESTRATE artifact |
Inputs (flags owned by the commands/grill.md shim)
- target — spec/plan path, a quoted topic, or empty (detect from context).
- --bound N — stop after N decision branches (quick gate; used by orchestrate Step 0.5).
- --no-capture — skip writing a
GRILL-*file; return decisions inline (embedded callers). - --yes / --non-interactive — auto-accept every Recommended answer, zero AskUserQuestion prompts.
- --refine / --no-refine — refine a topic via
prompt-refinerbefore grilling (default-on for topics; skipped for a path). - --no-tests / --no-docs — suppress the auto-emitted test-plan / Documentation sections.
Procedure
Step 1: Resolve the target
- Path argument → load that artifact (spec/plan/ORCHESTRATE/diff) and interrogate it.
- Quoted topic → sketch a 3–5 bullet skeleton, show it, then interrogate the skeleton.
- Empty → detect from
.STATUS, branch name, and recent commits; confirm the target before grilling.
Step 2: Codebase-first sweep (read-then-confirm)
When the argument is a quoted/bare topic, run prompt-refiner on it first (default-on;
--no-refine skips). When the argument is a path, skip refine entirely.
Before asking anything, explore the codebase — read repo evidence (.STATUS, git log,
docs/specs/*, commands/, ~/.claude/settings.json when relevant) — and PRE-ANSWER every
branch you can. Each pre-answer becomes the Recommended: line on its question — shown, never
silently skipped, so the user can override.
Step 3: Select attack angles (what to interrogate)
Grill is adversarial: interrogate the position's weaknesses, not its surface. Consider these generic angles and attack the ones that apply to THIS target, hardest first:
- Weakest recommendation — the call that looks worst in hindsight; name the cheaper/safer alternative.
- Riskiest assumption — the load-bearing belief that breaks the plan if false. Prioritise silent failures (behavior lost, not errors thrown).
- Implementation regret — the step you'd curse mid-build. Source project-specific traps at runtime from the target project's
CLAUDE.md/ ADRs / memory when aCLAUDE.mdis present in the target tree — do NOT hardcode any one project's gotchas here (that would poison grills of other projects). - Benefit honesty — is any claimed benefit (perf, token, time saved) measured or merely asserted? Separate real levers from wishful ones.
- Workflow discipline — does execution respect the project's documented workflow (branch model, test command, verify-before-merge) as stated in its
CLAUDE.md? - Blast radius — does the change silently alter a convention or require a doc/drift sweep? Scoped or hidden?
- Reversibility & scope creep — anything irreversible or ballooning past stated scope? Hold vs. proceed?
Angles are guidance for choosing branches, not a mandatory sweep — grill stays lightweight.
--bound N: pick the N highest-value angles for this specific target; a--bound 2gate still asks only 2 questions.- Unbounded: stop when the load-bearing branches are resolved — target roughly 5 branches, not an exhaustive tree. Milestone checkpoints (Step 4) let the user extend.
Step 4: The grill loop (deliberate, one question at a time)
Ask one question per AskUserQuestion call — one-at-a-time fidelity is the point; each answer reshapes the next question. This is a deliberate exception to craft's AskUserQuestion-batch convention (NOT 4-question batches). For EACH question:
- option[0] is the Recommended answer, labelled and reasoned;
- EVERY option carries a one-line consequence of choosing it;
- the implicit "Other" free-text path stays open for answers off the menu.
Halt on /done or empty-enter. --bound N stops after N branches. --yes /
--non-interactive: emit ZERO AskUserQuestion calls — auto-pick every Recommended, log each
pick, and proceed straight to capture. --yes cascades: the prompt-refiner auto-accepts AND the
interactive loop is suppressed — one flag, fully headless.
Milestone checkpoints (ADHD-friendly): every 5 resolved branches, pause with an AskUserQuestion: "keep going / wrap up now / show ledger so far" (reuses brainstorm's milestone pattern).
After each resolved branch: append it to the ledger immediately (Step 5 helper) so a crash/compaction never loses decisions.
Step 5: Capture (durable, own file, cross-linked)
If --no-capture (embedded callers like orchestrate Step 0.5): skip all file writes; return
the locked decisions inline to the caller. No GRILL-* file is created.
Otherwise grill writes its own file and never rewrites a brainstorm SPEC body:
from commands.utils.grill_ledger import (
resolve_ledger_path, spec_crosslink, add_backlink, append_decision)
path = resolve_ledger_path(topic, date, "docs/specs") # GRILL-<topic>-<date>.md
link = spec_crosslink(topic, "docs/specs") # latest brainstorm SPEC for this slug, or None
# write the GRILL header (include `link` if present), the decision ledger, and an ## Open Questions section
if link:
add_backlink(os.path.join("docs/specs", link), os.path.basename(path)) # idempotent, atomicEach resolved branch is appended live via append_decision. The brainstorm SPEC, if any, is only
ever touched by the single idempotent back-link.
Step 6: Handoff
Offer (one-directional into the planning spine):
/craft:plan tier 4 (plan-orchestrator) → ORCHESTRATE-*.md → /craft:do / /craft:orch.
Grill never executes — it interrogates and hands the locked artifact forward.
Do Not
- Reimplement grill logic in the
commands/grill.mdshim — this skill is the single source of truth. - Hardcode any project's specific gotchas into the attack angles (Step 3) — read the target's
CLAUDE.mdat runtime instead. - Rewrite a brainstorm SPEC's body — only the single idempotent back-link is permitted.
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.

