Technical FAQ

Common technical questions from the pilot team — based on FabriXWork recommended practices.

1. Preventing Auto-Installation

Q: Can we prevent users from automatically installing packages or tools?

A: Yes. FabriXWork has permission control files in place that allow administrators to manage what commands and installations users can execute.

How It Works

The app uses a permission configuration system that controls:

• Bash command execution
• Package installations (npm, pip, etc.)
• File editing capabilities

Example Permission Configuration

{
  "permission": {
    "bash": {
      "*": "ask",
      "npm install *": "deny",
      "pip install *": "deny"
    },
    "edit": "ask"
  }
}

Permission Levels

User-Type Recommendations

Based on pilot team feedback, here are recommended configurations by user type:

Important Notes

• No MDM required — FabriXWork has built-in permission controls
• Remote installation support — IT admins can help users install apps remotely (requires internal deployment alignment)
• Configuration balance — Strike a balance between security and usability
• Feedback loop — Regularly review and adjust permissions based on user needs

Note: A new feature is requested for centralized IT admin management of these permissions.

2. Safeguarding Confidential Data

Q: How does FabriXWork protect sensitive and confidential data?

A: FabriXWork is designed with data privacy as a core principle. Your files and conversations stay under your control.

Data Storage

AI Model Deployment Options

FabriXWork supports hybrid AI model deployment to meet different security requirements:

Option 1: Cloud AI Model

• How it works: Requests sent to cloud-based AI providers
• Best for: General business use, non-sensitive data
• Consideration: Ensure you trust the cloud provider

Option 2: Self-Hosted AI Model (On-Premise)

• How it works: Run AI models on your own infrastructure
• Best for: HR data, sensitive business information, regulated industries
• Consideration: Requires GPU resources and technical setup

Option 3: Self-Hosted AI Model (Cloud)

• How it works: Host AI models on your cloud infrastructure
• Best for: Teams needing scalability with control
• Consideration: Cloud hosting costs apply

Option 4: Local AI Model

• How it works: AI models run directly on user's machine
• Best for: Maximum privacy, offline scenarios
• Consideration: Limited by local hardware capabilities

PII Protection (Upcoming Feature)

An additional plugin is available for PII (Personally Identifiable Information) detection:

• What it does: Automatically detects sensitive information in your requests
• How it works: Replaces sensitive data with dummy data before sending to AI
• Use case: HR information, customer data, financial records

Recommendations by Data Type

Note: Some data "fragments" may still be sent to the AI model for processing. For highly sensitive data, use self-hosted options.

3. Controlling Application or System Updates

Q: How can we control when and how FabriXWork updates?

A: Updates can be managed through standard enterprise policies and private repositories.

Update Management Methods

Method 1: MDM Policy / Remote Install Server

• How it works: Use your existing Mobile Device Management (MDM) system
• Best for: Organizations with existing MDM infrastructure
• Control level: Full control over update timing and deployment

Method 2: Private Tool / Library Repository

• How it works: Use a private repository for tools and libraries
• Best for: Organizations with strict dependency control requirements
• Control level: Control which packages and versions are available

Method 3: Managed Repository (Selected Subscription Tier)

• How it works: FabriXWork provides a managed tool/library repository
• Best for: Teams without existing repository infrastructure
• Availability: Selected subscription tiers

Shortlisted Tool List

For organizations that want to pre-approve tools:

• BN (Business Needs) team can share a shortlisted tool list
• Only approved tools can be installed by users
• Reduces security risks from unvetted packages

Recommendation

4. Uninstalling Third-Party Programs

Q: What happens when we uninstall FabriXWork? Does it remove third-party programs it installed?

A: Currently, the uninstall command cleans FabriXWork files only. Third-party programs installed by agents remain on your system.

Current Behavior

Why?

Third-party programs may be used by other applications or workflows, so FabriXWork doesn't automatically remove them to avoid breaking other systems.

Upcoming Feature: Audit Trail

A new feature is in development:

• Audit logging of third-party programs — Track what programs/packages agents install
• Installation history — See when and why each program was installed
• Remediation support — Easier to identify and remove agent-installed programs

Best Practices

1. Review permission settings — Use deny for package installations if you want to prevent auto-installation
2. Regular audits — Periodically review installed programs on agent-used machines
3. Document installations — Keep a log of what agents have installed for your environment

5. Distributing Trained Agents

Q: Can we share custom-trained or configured agents between team members?

A: Export/Import functionality for agents is coming in the next Beta version.

Upcoming Feature: Agent Export/Import

Timeline: Next Beta version

Capabilities:

• Export agent configurations
• Import agents to other installations
• Share custom prompts and workflows
• Transfer agent knowledge bases

Availability

Current Workaround

Until export/import is available:

• Document agent configurations manually
• Share prompt templates via shared folders
• Use the Agent Directory to add the same agents across team members

Use Cases

Quick Reference

Related Resources

Have more questions? Contact the pilot program team or see Help & Troubleshooting.