KnowledgeDigest is engineered around local-first data integrity:

1. Zero Egress by Default: All document parsing, text chunking, and SQLite FTS5 search indexing occur 100% locally on your machine. No telemetry, usage statistics, or automated background telemetry calls are ever made.
2. Explicit LLM Boundary: External network calls only happen when the user explicitly configures and invokes LLM summarization (--flash via Google GenAI or --haiku via Anthropic).
3. Web Viewer Hardening:
   • State-changing endpoints (/open/, /api/reset_doc/, /api/delete_file/) strictly enforce HTTP POST (HTTP 405 on GET).
   • Host header validation mitigates DNS rebinding attacks.
   • Origin / Referer checks protect against Cross-Site Request Forgery (CSRF).
   • Search snippet outputs are escaped prior to <mark> tag injection to prevent stored XSS.
4. Unprivileged Execution: KnowledgeDigest runs as a standard user process (RunAsInvoker) without requiring elevated or root permissions.

If you discover a potential security vulnerability, please report it responsibly:

1. Do NOT open a public issue.
2. Use GitHub Private Vulnerability Reporting: Navigate to GitHub Advisory Submission.
3. Provide details:
   • Component / module affected (e.g. web_viewer.py, chunker.py, digest.py)
   • Steps to reproduce or proof-of-concept
   • Potential impact and severity assessment

• Initial Response: Within 48 hours of report submission.
• Triage & Assessment: Within 5 business days.
• Patch Deployment: Security fixes will be released on main and tagged in a timely manner before coordinated disclosure.