IntentKeeper takes security seriously, especially given our local-first architecture and browser extension integration.

If you discover a vulnerability in:

• Classification bypass (content that evades intent detection in a way that could cause harm)
• Browser extension security (XSS, injection, or privilege escalation in the extension)
• Data privacy (unintended data transmission, telemetry, or content exposure)
• Server-side vulnerabilities (code injection, path traversal, or API abuse)

Please report these privately by opening a private security advisory on GitHub.

Do NOT open a public issue for security vulnerabilities.

• Description of the vulnerability
• Steps to reproduce
• Potential impact
• Suggested fix (if you have one)

• Acknowledgment: Within 48 hours
• Initial assessment: Within 1 week
• Fix timeline: Depends on severity, but privacy and extension security issues are highest priority

• Browser extension vulnerabilities (XSS, content injection, CSP bypass)
• API server vulnerabilities (injection, SSRF, denial of service)
• Data exposure (classified content leaking, cache data accessible to other origins)
• Dependency vulnerabilities with a viable exploit path
• Bypass of the fail-open design that causes content to be incorrectly hidden or blocked

• Issues with Ollama itself (report to Ollama)
• Theoretical attacks requiring physical access to the user's machine
• Classification accuracy issues (these are feature requests, not security bugs)
• Social engineering attacks
• Issues already documented as limitations

For the full trust boundary, known gaps, and honest claims about what the architecture does and does not protect, see THREAT_MODEL.md.

IntentKeeper's security model is built on:

1. Local-first - All classification runs on user hardware via Ollama. No content is sent to external services.
2. Fail-open - If classification fails, content passes through unchanged. The system never silently blocks or hides content due to errors.
3. Minimal permissions - The browser extension requests only the permissions it needs.
4. No telemetry - Zero tracking, analytics, or usage data collection.

This project runs automated security checks on every push and PR:

• pip-audit - Dependency vulnerability scanning
• CodeQL - Static analysis for code vulnerabilities
• Gitleaks - Secret detection in commits

Thank you for helping keep IntentKeeper safe for everyone.