review-interface

Name: review-interface
Availability: InStock
Author: Miosa

Generate self-contained HTML review interfaces for JSONL data.

226|55|Updated Mar 19, 2026

Search Tags:#human-in-the-loop #jsonl #annotation #llm-evaluation #data-labeling #review-ui

Ready to run

npx skills add https://github.com/Miosa-osa/canopy --skill review-interface

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: review-interface
Source: https://github.com/Miosa-osa/canopy/tree/main/library/skills/workspace/review-interface
Command: npx skills add https://github.com/Miosa-osa/canopy --skill review-interface

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Annotating and auditing AI agent outputs and execution traces can be tedious and error-prone. This Skill generates self-contained HTML review interfaces that run entirely in the browser, enabling rapid, offline human evaluation of LLM outputs, agent traces, and labeled data.

Core Features & Use Cases

Create configurable annotation interfaces for reviewing, labeling, and comparing results from AI agents and models.
Supports loading data from JSONL, side-by-side comparisons, trace reviews, and export of labeled results for downstream QA and gold-standard dataset creation.
Works offline as a single self-contained HTML file with inline CSS/JS and keyboard navigation for efficient human-in-the-loop workflows.

Quick Start

Run /review-interface with your data file to generate a self-contained HTML review UI.

Frequently Asked Questions about review-interface

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

FAQPage Schema

How do I create a human-in-the-loop annotation UI for reviewing LLM agent traces?▼

To create a human-in-the-loop annotation UI, analyze your JSONL data to generate a self-contained HTML interface. This offline browser tool enables rapid labeling and auditing of LLM outputs with configurable schemas and keyboard navigation.

What is the best way to label LLM evaluation datasets offline?▼

The best way to label LLM evaluation datasets offline is generating a single self-contained HTML file. It runs entirely in the browser with inline CSS/JS, supporting configurable labels, blind mode to prevent bias, and result exports.

Can I review and compare AI agent outputs side-by-side without an internet connection?▼

Yes, you can review and compare AI agent outputs offline. The generator produces a self-contained HTML file with inline assets, enabling side-by-side comparisons and trace reviews directly in the browser without any server or internet connection.

Does the review interface support blind mode to prevent metadata bias during data labeling?▼

Yes, the review interface supports blind mode to prevent metadata bias during data labeling. This optional feature hides metadata within the self-contained HTML UI, ensuring objective human evaluation of LLM outputs and agent execution traces.

How do I export labeled results from a JSONL file for downstream QA?▼

You can export labeled results for downstream QA directly from the self-contained HTML interface. After loading your JSONL data and completing annotations using the configurable schemas, the UI provides tools to export the labeled dataset for gold-standard creation.