craft-goal

Creates concise goal prompts with verifiable conditions for autonomous task execution.

Updated May 16, 2026
One-click install
npx skills add https://github.com/ckyeon/claude-code-settings --skill craft-goal
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: craft-goal
Source: https://github.com/ckyeon/claude-code-settings/tree/main/user/shared/skills/craft-goal
Command: npx skills add https://github.com/ckyeon/claude-code-settings --skill craft-goal

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill solves the problem of creating detailed, yet concise, goal prompts that a separate evaluator can use to assess the success of a task. It helps users craft prompts that are suitable for autonomous execution and can be judged purely from the conversation transcript.

Core Features & Use Cases

  • Prompt Composition: Assembles goal conditions from user descriptions to guide autonomous task execution.
  • Verification & Constraints: Ensures that conditions are verifiable by commands and includes explicit constraints to prevent unnecessary changes.
  • Progress Tracking & Turn Limits: Includes optional progress tracking for long tasks and turn limits for safety and bounded execution.

Quick Start

Create a goal prompt using the craft-goal skill by providing a detailed task description. For example, "Generate a goal prompt to refactor the 'src/utils' module while ensuring no public API changes."

Frequently Asked Questions about craft-goal

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

FAQPage Schema
How do I create goal prompts for autonomous task execution?

To create goal prompts for autonomous task execution, you provide a detailed task description. The system then assembles clear conditions and explicit constraints to guide the agent without requiring direct execution for verification.

What is an executable goal prompt and when do I need one?

An executable goal prompt is a concise instruction set that an evaluator uses to assess task success purely from a transcript. You need one when automating tasks that require unambiguous, condition-based verification.

How do I prevent prompt injection during autonomous execution?

To prevent prompt injection during autonomous execution, you must craft goal prompts with explicit constraints. This ensures the system avoids unnecessary changes and bounds execution safely within defined limits.

Can I set turn limits and progress tracking for long autonomous tasks?

Yes, you can set turn limits and optional progress tracking for long autonomous tasks. These features ensure bounded execution and provide safety by limiting the scope of automated operations.

How do I ensure my autonomous task conditions are verifiable without executing code?

To ensure task conditions are verifiable without executing code, the goal prompt must include condition statements verifiable by commands. This allows an evaluator to judge success purely from the conversation transcript.

What is the best way to structure a goal prompt for a complex refactoring task?

The best way to structure a goal prompt for a complex refactoring task is to assemble verifiable conditions from your description, add explicit constraints to prevent unwanted changes, and include turn limits for safety.