craft-ice

Converts an epic or feature description into an ICE specification with Intent, Context, Expectation.

3|Updated Jan 25, 2026
One-click install
npx skills add https://github.com/kapilvirenahuja/garura --skill craft-ice
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: craft-ice
Source: https://github.com/kapilvirenahuja/garura/tree/main/core/components/plays/craft-ice
Command: npx skills add https://github.com/kapilvirenahuja/garura --skill craft-ice

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

It converts an epic or feature description into a complete, downstream-ready specification so teams can proceed without repeatedly re-deriving requirements and success criteria.

Core Features & Use Cases

  • ICE-compliant spec generation: Produces a clean Intent triple, a grounded Context bundle, and a vetted Expectation with success scenarios and recovery.
  • Substantial-work scope enforcement: Refuses fix-it-scale changes and redirects to the appropriate quick-fix flow with evidence.
  • Human-in-the-loop checkpointing: Applies configurable review gating before downstream plays can run.

Quick Start

Run craft-ice on the current epic or feature branch and answer the discovery questions to produce an approved Intent/Context/Expectation spec for downstream prepare or enhance.

Frequently Asked Questions about craft-ice

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

FAQPage Schema
How do I convert an epic into a specification with requirements and success criteria?

To convert an epic into a specification, the skill generates an implementation-agnostic Intent triple, a grounded Context bundle, and a vetted Expectation with success and recovery scenarios for downstream consumption.

What is intent-driven development and how does it assemble context for engineering delivery?

Intent-driven development assembles requirements once by generating an Intent triple and a Context bundle grounded in memory and codebase, ensuring downstream teams reuse vetted specifications without re-deriving success criteria.

How do I generate testable outcomes for enterprise feature delivery?

You generate testable outcomes by producing a vetted Expectation component that defines success scenarios and recovery criteria, which undergoes human-in-the-loop review gating before downstream plays run.

Can I use this specification approach for small bug fixes and minor changes?

No, this specification approach refuses fix-it-scale changes and redirects them to the appropriate quick-fix flow with evidence, as it applies only to substantial engineering work requiring context grounding.

How do I configure human-in-the-loop review checkpoints before downstream execution?

You configure human-in-the-loop review checkpoints by applying configurable review gating within the skill, which blocks downstream prepare or enhance plays until the Expectation spec receives explicit approval.