grill-with-docs

Interrogates a plan or design through structured questions while generating ADRs and glossary entries.

Updated Jul 29, 2026
One-click install
npx skills add https://github.com/MaiconGambini/opencode-harness-guide --skill grill-with-docs-maicongambini
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: grill-with-docs
Source: https://github.com/MaiconGambini/opencode-harness-guide/tree/main/skills/grill-with-docs
Command: npx skills add https://github.com/MaiconGambini/opencode-harness-guide --skill grill-with-docs-maicongambini

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Plans and designs often contain hidden assumptions, unresolved ambiguities, and undocumented decisions that surface too late. This Skill runs a relentless question-by-question interview that sharpens a plan while simultaneously producing durable documentation (ADRs and a glossary) as a byproduct. ## Core Features & Use Cases - HITL interview mode: Interviews the operator live, one question at a time, walking the decision tree of a plan or design. - Autonomous AFK mode: When driven by a scheduler (e.g. spec-lead in an authorized run), it answers each question itself from the codebase, existing ADRs, CONTEXT.md, and stated requirements, labeling assumption-based answers explicitly. - Documentation as output: Writes Architecture Decision Records and glossary entries via the /domain-modeling skill in both modes. - Use Case: A tech lead has a draft design for a new service. Run a grilling session to expose every unresolved decision, record each resolution as an ADR, and hand reviewers a plan with assumptions clearly marked. ## Quick Start Run a grilling session on my current design plan to challenge every decision and produce ADRs and glossary entries as we go.

Frequently Asked Questions about grill-with-docs

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I stress-test a design plan before implementation?

Run a grilling session that walks the decision tree of your plan one question at a time. Each resolved question is recorded as a decision, and Architecture Decision Records plus glossary entries are written as the interview progresses.

What is the difference between HITL and AUTO mode in a grilling session?

HITL mode interviews a live operator one question at a time, while AUTO mode runs self-driven without pausing, answering each question from the codebase, existing ADRs, CONTEXT.md, and requirements. AUTO labels assumption-based answers so reviewers can audit them.

Can a scheduler run a design review without a human present?

Yes, AUTO (AFK) mode supports scheduler-driven runs such as spec-lead inside an authorized run. It escalates to the operator only on genuine blockers like unresolvable ambiguity, irreversible decisions without a defensible default, or scope changes.

What documentation does a grilling session produce?

Each session produces Architecture Decision Records and glossary entries written through the domain-modeling skill. The documentation output is identical in both HITL and AUTO modes; only the source of the answers changes.

When should I not use autonomous mode for plan review?

Avoid AUTO mode when decisions are hard to reverse and lack a defensible default, or when requirements are too ambiguous for any available source to settle. Those cases require live operator judgment and trigger escalation even in autonomous runs.