What is the primary purpose of the feedback-ls skill?

The feedback-ls skill is designed to send document pages and transcriptions to Label Studio, allowing humans to review, correct, and provide feedback on AI-generated text or segmentation.

Is there a way to test the import before sending data to Label Studio?

Yes, the tool includes a dry_run parameter that allows you to see the converted Label Studio JSON structure without actually performing the import.

How are the Label Studio credentials managed?

The skill uses environment variables (LS_URL, LS_TOKEN, and LS_PROJECT_ID) to securely handle connection details, though these can be overridden if necessary.

Do I need existing AI transcriptions to use this tool?

No. While it can import ALTO XML for pre-annotation, it also features an 'Images only' mode for creating blank tasks meant for manual annotation from scratch.

Can I assign tasks to specific team members automatically?

Yes, by providing a user's email in the assign_to parameter, the imported tasks will be automatically assigned to that specific reviewer in Label Studio.

Label Studio Feedback Integration

Name: Label Studio Feedback Integration
Author: AI-Riksarkivet

byAI-Riksarkivet

•

データサイエンスとML

Streamlines the process of importing document pages and transcriptions into Label Studio for manual review and human-in-the-loop feedback.

This skill enables Claude to facilitate human-in-the-loop workflows by exporting archival document images and AI-generated transcriptions (ALTO XML) directly into Label Studio projects. It supports two primary modes: a pre-annotated import that maps text line polygons and existing transcriptions for correction, and an image-only mode for starting manual annotations from scratch. By automating the technical overhead of task creation and data mapping, it bridges the gap between automated AI processing and high-quality human verification, making it an essential tool for archival digitization, HTR training, and rigorous quality assurance.

主な機能

01Dry run functionality to preview JSON payloads before project import

02Automated import of images and ALTO XML transcriptions into Label Studio

038 GitHub stars

04Pre-annotation support including text line polygons and VectorLabels

05Image-only mode for manual annotation projects without existing data

06Direct task assignment to specific reviewers via email mapping

ユースケース

01Managing quality assurance workflows for large-scale historical document digitization

02Creating high-quality ground truth datasets for training HTR and OCR models

03Flagging AI-generated transcriptions for manual correction by subject matter experts

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add ai-riksarkivet/ra-mcp feedback-ls

For use in Claude.ai and ChatGPT

Download Skill