Ollama Dev Companion takes security seriously. This document outlines our security practices, vulnerability reporting process, and security considerations for users and developers.

If you discover a security vulnerability, please follow these steps:

1. DO NOT create a public issue
2. Email security concerns to: [maintainer email]
3. Include:
   • Description of the vulnerability
   • Steps to reproduce
   • Potential impact
   • Suggested fix (if any)

• Acknowledgment: Within 48 hours
• Initial Assessment: Within 1 week
• Fix Development: Depends on severity
• Public Disclosure: After fix is released

All user inputs are validated to prevent:

• Command Injection: Sanitized before execution
• Path Traversal: Restricted to workspace boundaries
• Code Injection: Escaped in webviews
• Size Limits: Enforced on all inputs

// Example validation
const validatedPath = validationService.validateFilePath(userPath);
if (!validatedPath.isValid) {
  throw new ValidationError(validatedPath.error);
}

• Validates all file paths
• Prevents access outside workspace
• Resolves symbolic links
• Logs all file operations

// File access is always validated
const result = await secureFileAccess.readFile(path);
if (!result.success) {
  // Access denied
}

• Canonical path resolution
• Workspace boundary enforcement
• Symbolic link detection
• Hidden file protection

• Local-only by default (localhost:11434)
• No external API calls without user consent
• Configurable timeout limits
• Request size limits

• Currently relies on local Ollama instance
• No credentials stored
• No cloud services used

const csp = [
  `default-src 'none'`,
  `style-src ${webview.cspSource} 'unsafe-inline'`,
  `script-src 'nonce-${nonce}'`,
  `img-src ${webview.cspSource} https: data:`,
  `font-src ${webview.cspSource}`
].join('; ');

• Nonce-based script execution
• No eval() or inline scripts
• Restricted resource loading
• Message validation

• No telemetry collection
• No usage analytics
• No cloud storage
• No external API calls (except Ollama)
• No credential storage

• All data stays on user's machine
• Chat history stored locally
• Cache stored in workspace
• Settings in VS Code configuration

• System prompts are isolated
• User input is clearly delineated
• No execution of generated code
• Clear marking of AI-generated content

• HTML escaping in webviews
• Markdown rendering restrictions
• No script execution from AI output

1. Keep Ollama Updated: Use latest Ollama version
2. Local Models Only: Don't expose Ollama to network
3. Review Generated Code: Always review AI suggestions
4. Workspace Isolation: Use separate workspaces for sensitive projects
5. Extension Updates: Keep extension updated

1. Input Validation: Always validate user input
2. Use Type Guards: Runtime type checking
3. Dependency Injection: Use DI for testability
4. Error Handling: Don't expose internal errors
5. Logging: Log security events

• All inputs validated
• File paths checked against workspace
• No use of eval() or Function()
• CSP implemented in webviews
• Errors don't leak sensitive info
• Dependencies up to date

• Security scan completed
• Dependencies audited
• No debug code in release
• Permissions minimized
• Documentation updated

• Models run locally via Ollama
• Model behavior depends on training
• No guarantees on model output safety

• Extension has file system access
• Limited to workspace by default
• User can grant broader access

• Extension doesn't execute generated code
• User responsible for reviewing suggestions
• No automatic code execution

# TypeScript strict mode
npm run build

# ESLint security rules
npm run lint

# Dependency audit
npm audit

• Input validation service
• Secure file access service
• Rate limiting
• Memory monitoring

1. Isolate: Disable affected functionality
2. Assess: Determine scope and impact
3. Fix: Develop and test patch
4. Release: Deploy fix quickly
5. Disclose: Inform users appropriately

Security fixes are released as:

• Critical: Immediate patch release
• High: Within 1 week
• Medium: Within 2 weeks
• Low: Next regular release

• Weekly automated dependency scans
• Manual review of critical updates
• Automated PR for updates

• ollama: Official Ollama client
• vscode: VS Code extension API
• Build tools (dev dependencies only)

• Follows VS Code security guidelines
• Regular marketplace security scans
• Permission declarations accurate

• MIT licensed
• Security issues tracked publicly (after fix)
• Community security contributions welcome

For security concerns:

• Email: [security contact]
• GitHub Security Advisories
• Private vulnerability reporting

We thank security researchers who responsibly disclose vulnerabilities and help make Ollama Dev Companion more secure.