workflow-grill

Resolve ambiguous requirements via prioritized decision trees with repository evidence.

2|8|Updated Apr 20, 2026
One-click install
npx skills add https://github.com/lugassawan/swe-workbench --skill workflow-grill
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: workflow-grill
Source: https://github.com/lugassawan/swe-workbench/tree/main/skills/workflow-grill
Command: npx skills add https://github.com/lugassawan/swe-workbench --skill workflow-grill

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It resolves under-specified requirements by forcing a disciplined, evidence-first interrogation loop before any implementation or delegation begins.

Core Features & Use Cases

  • Decision-tree clarification: Turns ambiguous intent into a prioritized set of decisions that must be resolved to produce correct artifacts.
  • Codebase-first answers: Searches the repository first and uses precise file:line evidence to resolve questions without asking the user when the code is clear.
  • Single-step prompting: Asks only one focused question per turn, with a recommended answer, until exit conditions are met.
  • Handback as Resolved decisions: Outputs exactly one fenced Resolved decisions block that commands thread into their normal artifact/delegation step.

Quick Start

Activate workflow-grill by running a command such as /capture with a sufficiently described objective, then answer the one-question-at-a-time prompts until the Resolved decisions block is produced.

Frequently Asked Questions about workflow-grill

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

FAQPage Schema
How do I clarify ambiguous requirements before starting code implementation?

You can resolve ambiguous requirements by walking a prioritized decision tree one question at a time, grounding answers in repository evidence using file:line citations before implementation begins.

What is the best way to interrogate codebase decisions without asking the user unnecessary questions?

The best way to interrogate codebase decisions is searching the repository first and using precise path/to/file:line evidence to resolve questions automatically, only prompting the user when the code is unclear.

How does a decision tree clarification loop work for interactive software engineering commands?

The decision tree clarification loop applies to interactive commands like capture, design, implement, architect, extend, and debug by asking one focused question per turn with a recommended answer until strict exit conditions are met.

Can I use orchestrator mode to resolve under-specified requirements for software design?

Yes, you can use orchestrator-only operation to resolve under-specified requirements for software design by forcing a disciplined, evidence-first interrogation loop before any implementation or delegation begins.

How to thread resolved decisions into artifact generation after requirements clarification?

To thread resolved decisions into artifact generation, output exactly one fenced Resolved decisions block that downstream commands thread into their normal artifact or delegation step.

When should I not use an evidence-first interrogation loop for requirements clarification?

You should not use an evidence-first interrogation loop as a standalone from-scratch workflow, as it requires orchestrator-only operation and is designed to clarify steps inside interactive commands rather than running independently.