agentsclimarketplace

Grill

Skill andresnator/agents-orchestrator/skills/grill

Trigger: grill me, grill with docs, grill me sdd, grill code, grill docs open, entrevistame, entrevistame sdd. Router for relentless interview modes: plain, docs, or SDD planning.From its SKILL.md

Install
npx -y skills add andresnator/agents-orchestrator --skill grill

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its file declares

Copied from the file, not written here

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

4.4 KB, ~1.0k tokens by cl100k_base, as published. Nobody here has run it

Activation Contract

Use when the user wants a relentless interview for a plan, design, code approach, docs-backed exploration, or SDD planning set.

Select exactly one mode:

Trigger shapeModeBehavior
grill me, grill code, entrevistameplainRun grilling with native question UX.
grill with docs, grill docs opendocsRun grilling plus domain-modeling with native question UX.
grill me sdd, entrevistame sdd, SDD planning interviewsddInterview, then draft OpenSpec proposal.md, specs, design.md, and tasks.md.

Hard Rules

  • Every user-facing question MUST go through the native-question-ux skill. Use its fallback if the runtime has no native question tool.
  • Follow the grilling skill in every mode: one question at a time, attach a recommendation, stop and wait, and explore the codebase instead of asking when possible.
  • Preserve open-ended interview semantics; do not turn the flow into generic code review.
  • docs mode also follows domain-modeling to capture terms, decisions, and context as they emerge.
  • Interview/summaries/gates use the user's language; artifacts default to English unless the user explicitly requests Spanish artifacts.
  • sdd mode is plan-only: read-only codebase access; no code edits, builds, installs, tests, commits, or state-changing commands. The only side effect is writing approved planning .md files.
  • In sdd mode, subskills return approved drafts only; this orchestrator owns the single write step.

Plain Mode

  1. Load grilling.
  2. Ask each interview question through native-question-ux.
  3. Stop after each question and wait for the answer.
  4. Return the next question, recommendation, or conclusion from the active grilling flow.

Docs Mode

  1. Load grilling and domain-modeling.
  2. Ask each interview question through native-question-ux.
  3. Capture useful terminology, decisions, and context through the domain-modeling flow when the user approves documentation.
  4. Stop after each question and wait for the answer.
  5. Return the next question, recommendation, captured-doc status, or conclusion from the active flow.

Decision Gates

SituationAction
Missing change namePropose a kebab-case, verb-led name and confirm.
.ai/orchestrator/changes/{change-name}/ existsAsk resume vs rename.
Product/business uncertaintyRoute to sdd-draft-proposal.
Observable behavior uncertaintyRoute to sdd-draft-spec.
Technical/architecture uncertaintyRoute to sdd-draft-design.
Slicing or task-ordering uncertaintyRoute to sdd-draft-tasks.
User stops earlyOffer to write approved artifacts only.

Execution Steps

Use these steps only in sdd mode:

  1. Set language and confirm the change name.
  2. If an existing change folder exists, ask whether to resume or rename.
  3. Run sdd-draft-proposal; loop inline until the proposal draft is approved.
  4. Build the capability work list from proposal Capabilities.
  5. Ask once whether spec approval should be per-file or batched.
  6. Run sdd-draft-spec for each capability or batch; loop until approved.
  7. Run sdd-draft-design; it must silently explore the real codebase read-only before technical questions or draft.
  8. Run sdd-draft-tasks with the approved spec and design; loop inline until the tasks draft is approved.
  9. Ask once to write N files under .ai/orchestrator/changes/{change-name}/.
  10. On yes, write exactly proposal.md, specs/{capability}/spec.md, design.md, and tasks.md; on no, leave drafts inline.

Output Contract

Return approved drafts, write status, paths written or pending, and unresolved questions. If a referenced skill cannot be resolved by name in the current runtime, say so instead of silently continuing.

References

  • grilling skill
  • native-question-ux skill
  • domain-modeling skill
  • sdd-draft-proposal skill
  • sdd-draft-spec skill
  • sdd-draft-design skill
  • sdd-draft-tasks skill

Attribution

Inspired by Matt Pocock's grilling skills at https://github.com/mattpocock/skills; adapted for SDD planning orchestration.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Gives 2 of the 12 instructions most review quality skills give in ~1.0k tokens

Counted across 1,048 of the 1,783 authors here whose files we hold, read 2026-08-07

  • Ask questions one at a timehere, and in 81 of 1048, across 64 files
  • Provide a recommended answer for each questionin 73 of 1048, across 50 files
  • Explore the codebase instead of asking answerable questionshere, and in 66 of 1048, across 42 files
  • Resolve dependencies between decisions one-by-onein 42 of 1048, across 17 files
  • Interview the user relentlessly about the planin 38 of 1048, across 13 files
  • Order findings by severityin 31 of 1048
  • Resolve each branch of the decision treein 27 of 1048, across 5 files
  • Run a grilling sessionin 26 of 1048, across 5 files
  • Update CONTEXT.md immediately when a term is resolvedin 26 of 1048, across 11 files
  • Propose precise canonical terms for vague languagein 25 of 1048, across 7 files
  • Create documentation files lazilyin 24 of 1048, across 5 files
  • Assign severity to every findingin 24 of 1048

Said here and by no other author read

  • Select exactly one interview mode per request
  • Route every user-facing question through native-question-ux
  • Attach a recommendation to each interview question
  • Capture domain terms and decisions during docs mode
  • Conduct interviews and gates in the user's language
  • Write approved planning artifacts in English unless requested otherwise

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 326,286. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.