Why does it require 2-4 options for every question?

Providing specific, labeled options with implications prevents open-ended questions and helps the developer make an informed decision quickly based on analyzed trade-offs.

When does the User Input Protocol trigger?

It triggers when Claude encounters business rule ambiguities, architecture trade-offs, scope questions, or destructive actions that require human judgment.

Does this protocol stop all work immediately?

It stops the specific decision path requiring input. While it prevents making assumptions on that path, Claude can still proceed with other unrelated work if possible.

How does it handle background subagents?

It saves the subagent's state (task status, analysis, and modified files) before pausing, allowing it to resume instantly without re-doing work once the user responds.

What is the Factory Mode urgency classification?

It labels decisions as blocking, next-review, or informational to help manage human attention in automated CI/CD pipelines or batch processing workflows.

User Input Protocol

Name: User Input Protocol
Author: jwilger

byjwilger

•

生产力与工作流

Implements a structured checkpoint system for AI agents to request human decisions without making assumptions.

The User Input Protocol skill provides a robust framework for Claude to handle ambiguity by pausing at critical decision points rather than guessing. It defines a standardized AWAITING_USER_INPUT format that presents context, multiple labeled options with their implications, and agent recommendations. This skill is essential for managing architectural trade-offs, business rule conflicts, and destructive actions, ensuring the developer remains in control. It includes advanced patterns for subagent state preservation, allowing background tasks to save progress and resume instantly once a decision is provided, and offers urgency classification for automated pipelines.

主要功能

012 GitHub stars

02State preservation and resumption patterns for subagents and background tasks

03Prescriptive options (2-4) with detailed implications instead of open-ended questions

04Urgency classification (blocking, next-review, informational) for factory workflows

05Decision batching to group related questions and reduce developer interruption

06Structured AWAITING_USER_INPUT checkpoint format for clear human communication

使用场景

01Gaining explicit confirmation before performing destructive actions like file deletion

02Resolving architectural trade-offs where multiple valid implementation paths exist

03Clarifying ambiguous business requirements during complex feature development

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jwilger/agent-skills user-input-protocol

For use in Claude.ai and ChatGPT

Download Skill