Key Concepts

In KroWork, we have built a flexible, secure and sustainable low-code collaboration system based on five core concepts. Understanding these concepts will help you grasp how KroWork assists you in converting fragmented workflows into reusable and manageable digital assets.

I. Task — Dialogue-Driven Autonomous Work Unit

A task refers to a complete requirement you submit to KroWork through natural language. Simply chat with KroWork in natural language, and it will help you:

• Create local desktop applications and deploy automated office workflows
• Organize file data, perform batch operations and conduct in-depth research
• Write scripts, design interfaces and run various automated tasks

II. Skill — Atomic Capability Library Schedulable by Tasks

Skills are standardized built-in capability modules of KroWork, serving as the fundamental atomic capabilities for building various applications. The main skills are listed below.

• Interface Building: Quickly build interactive desktop interfaces.
• Browser & Automation: Two ways to automate the web:
  • Browser Bridge controls your real browser via a Chrome extension.
  • Browser Use runs an independent browser, no extension needed.
• Data & Analytics: Clean, process, and visualize data. Handles structured datasets, statistical charts, deduplication, error correction, and standardization.
• Reports & Writing: Generate reports and export them as PPTX or PDF in one click.
• Research & Information: Run deep research across multiple sources, track industry trends, and pull live information from the web.
• App Patterns: Building blocks for Kro Apps: scheduled tasks, pop-up reminders, linked components, and recurring web scraping.
• File Management: Automatically sort, rename, and archive files and folders.

III. Kro App Installation — Adding a Kro to Your Desktop

Kro App Installation is a core feature of KroWork designed to boost your productivity.

Once you complete a workflow, you can install it directly as a locally runnable, independent Kro App for one-click reuse later. This eliminates the need for repeated conversations and reduces unnecessary token consumption, turning one-off tasks into dedicated, long-term tools that you can launch directly from the macOS Launchpad / Windows Start Menu for instant access.

IV. Sandbox — Isolated Execution Security Buffer Layer

The sandbox is an isolated runtime environment that ensures operational security for KroWork. All potentially influential actions (code execution, web browsing, file modification, etc.) are first executed in the sandbox. Changes will only be written to the actual system upon your confirmation.

• Default isolation: Temporary files, code execution results and file modification operations generated by any skill invocation all take place in the temporary sandbox space.
• Change confirmation: When modifications in the sandbox need to be applied to the real file system, KroWork will explicitly notify you of the files to be altered and will only write changes after your confirmation.
• Risk preview: You can preview execution outcomes in the sandbox first (e.g., batch renaming preview, deletion simulation) before implementation.
• Side-effect-free trial and error: Unconfirmed modifications will be automatically discarded, and the sandbox environment can be reset at any time to prevent losses caused by misoperations.

V. Permission — Explicit Control with Minimum Authorization

KroWork strictly adheres to the principle of least privilege, requiring explicit user authorization for all sensitive operations. The system offers three permission approval modes to suit different usage scenarios, balancing security and efficiency:

Default Mode

• Core Logic: All operations involving risks or exceeding the defined usage scope require manual user approval before execution.
• Use Cases: First-time usage, handling sensitive data, or scenarios with extremely high requirements for operational security.
• Features: You personally oversee every high-risk step, maintaining clear control over every capability call and eliminating risks caused by misoperation at the source.

Auto Mode

• Core Logic: The AI automatically identifies safe and compliant routine operations, authorizing their execution without manual confirmation; only high-risk operations trigger approval requests.
• Use Cases: For experienced users, running highly repetitive automated tasks, or scenarios where you want to reduce frequent pop-up interruptions.
• Features: Balances efficiency and security. Safe operations are allowed with one click, while high-risk actions are still proactively blocked, ensuring a smooth experience without compromising data safety.

Full Allow Mode

• Core Logic: Allows most routine operations automatically. Only blocks high-risk destructive actions like deleting system files or formatting disks.
• Use Cases: Batch automation processing, handling non-sensitive local data, or professional scenarios requiring maximum operational efficiency.
• Features: Almost no approval pop-ups, ideal for fully controlled local automated workflows. It retains a final block on destructive operations to prevent irreversible system risks.