zacczakk/metronome/grill-with-docs

Claude Code and agent skill by zacczakk/metronome

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Use when a plan, design, requirement, or implementation approach needs stress-testing before code, especially when project language, hidden constraints, docs, or architectural decisions matter.

Install

npx vskill@latest install zacczakk/metronome/grill-with-docs

SKILL.md

Grill With Docs

Overview

Shared understanding beats fast misunderstanding. Interview the user until the plan is crisp, but do not ask questions the repository can answer.

When to Use

  • User asks to be grilled, challenged, stress-tested, or aligned before implementation.
  • Requirements are fuzzy, high-risk, or likely to encode hidden domain language.
  • The change may create durable terminology, scope boundaries, or hard-to-reverse decisions.
  • Before writing a PRD, plan, or agent brief when the current context is incomplete.

Do not use for trivial edits where the desired change is already concrete and reversible.

Core Loop

  1. Read repo guidance first: AGENTS.md, README.md, relevant docs/, existing docs/context.md, and docs/adr/ if present.
  2. Explore code when it can answer a question. Never ask the user to restate facts visible in code or docs.
  3. Ask one question at a time. Each question should resolve a real branch of the design tree.
  4. Include your recommended answer with each question.
  5. When a term or decision crystallizes, capture it immediately if it will help future agents.
  6. Continue until scope, language, constraints, acceptance criteria, and open risks are explicit.

Context Doc

Use docs/context.md as the canonical shared-language file. Create it lazily only after the first useful term or ambiguity is resolved.

md
# Context
## Language
**Canonical Term**: One-sentence definition.
_Avoid_: Ambiguous synonym, stale name
## Relationships
-**Term A** owns **Term B**.
## Flagged Ambiguities
-"account" used for both **Customer** and **User**; resolved as distinct concepts.

Rules:

  • Add only domain/project concepts, not general programming terms.
  • Pick one canonical term and list aliases to avoid.
  • Keep definitions one sentence.
  • Record relationships when they clarify boundaries.

ADRs

Use docs/adr/NNNN-short-title.md. Create docs/adr/ lazily.

Offer an ADR only when all are true:

  • Hard to reverse.
  • Surprising without context.
  • Result of a real tradeoff.

ADR body can be tiny:

md
# Short Decision Title
We chose X over Y because Z. This matters because future maintainers might otherwise assume A.

Question Quality

Good questions are specific and actionable:

  • "Should the import pipeline treat duplicate rows as idempotent updates or validation errors? Recommended: idempotent updates, because reruns should be safe."
  • "Your docs define Customer as the billing owner, but this flow talks about end users. Are those distinct? Recommended: yes, use User for login identity."

Bad questions:

  • "Can you provide more context?"
  • "What should I know?"
  • "How should this work?"

Stop Condition

Stop when you can state:

  • Problem and desired outcome.
  • In-scope and out-of-scope work.
  • Canonical terms and resolved ambiguities.
  • Key constraints and tradeoffs.
  • Acceptance criteria or next artifact to produce.

Then summarize and recommend the next skill or artifact: agent-brief, writing-plans, oe-create-plan, or implementation.

Common Mistakes

MistakeFix
Asking before reading docsRead first, then ask only unresolved questions.
Batching questionsOne question at a time.
No recommendationInclude your recommended answer.
Capturing every thoughtCapture only durable terms and hard decisions.
Creating ADRs for obvious choicesADR only when hard to reverse, surprising, and tradeoff-driven.

Security

  • Security scanPassed, 98 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
0

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/zacczakk/metronome/grill-with-docs/badge)](https://verified-skill.com/skills/zacczakk/metronome/grill-with-docs)

Something wrong with this skill or its page? Tell us.