What is the purpose of the Feedback Handler skill?

It manages human-in-the-loop (HITL) interactions, allowing AI workflows to pause and wait for human input or approval before continuing to the next phase.

Can I customize the options presented to the user?

Yes, the skill supports multiple feedback types including approval, selection, and review, which allow for custom or predefined response options.

How does this skill notify users that feedback is needed?

The skill can post formatted requests directly to GitHub issue comments or provide interactive prompts within the command-line interface (CLI).

How does a workflow resume after feedback is provided?

Once feedback is received, the skill validates the response, updates the run state to 'in_progress', and emits a signal to the manager to resume from the specific step.

Does it track the history of decisions made during a run?

Yes, every response is processed and stored in a feedback history array within the workflow state file for auditing and debugging purposes.

Faber Feedback Handler

Name: Faber Feedback Handler
Author: fractary

$fractary$ byfractary

•

协作工具

Manages human-in-the-loop feedback requests for FABER workflows by posting to issues and tracking response states.

The Faber Feedback Handler skill enables seamless human interaction within automated AI workflows by managing decision points and feedback loops. It automates the process of requesting approvals, clarifications, or design reviews by posting formatted comments to GitHub issues and tracking the state of these requests within the FABER ecosystem. This ensures that long-running tasks remain organized, providing developers with clear context and actionable options to resume or pivot automated operations safely without manual state management.

主要功能

01Automated feedback requests via GitHub issue comments

02Automated resume-point management for workflow continuation

03Standardized feedback templates for approvals, reviews, and error resolutions

04Real-time state tracking for Human-In-The-Loop (HITL) decision points

05Integration with CLI prompts and workflow event emitters

062 GitHub stars

使用场景

01Managing error resolution by asking developers to retry, skip, or abort failed steps

02Requiring manual architectural approval before proceeding to implementation

03Collecting specific technical clarifications during complex code refactoring

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add fractary/claude-plugins feedback-handler

For use in Claude.ai and ChatGPT

Download Skill