Skip to content

Security: ellmos-ai/hungrycall

Security

SECURITY.md

Security Policy / Sicherheitsrichtlinie

Reporting a Vulnerability / Sicherheitslücke melden

If you discover a security vulnerability or security concern within any repository in the ellmos-ai organization, please report it responsibly:

  1. Do NOT open a public issue or disclose vulnerability details publicly before a fix is available.
  2. Use GitHub Security Advisories on the affected repository to create a private draft advisory.
  3. Or contact the maintainers directly via email:
    • security@ellmos.ai
    • security@open-bricks.org
    • lukas@open-bricks.org
    • support@lukasgeiger.com

Response Timeline / Reaktionszeit

  • Acknowledgment: Within 48 hours (best effort, guaranteed within 7 days)
  • Initial Assessment & Triage: Within 7 to 14 days
  • Fix & Disclosure Coordination: Best effort, typically within 30 days depending on severity

Supported Versions / Unterstützte Versionen

Repository / Tool Supported Release Security Updates
Organization profile (ellmos-ai/.github) Latest commit on master ✅ Supported
Active public repositories (bach, rinnsal, open-ocean, usmc, gardener, skills, MCP servers, agent operations tooling, domain tools, media workflows) Latest commit or latest stable tag on default branch (master/main) ✅ Supported
Archived legacy repositories (recludos-legacy) Archived read-only ❌ No active updates

Security Invariants / Sicherheitsinvarianten

  • Zero-Egress & Local-First: All LLM operating system components, memory stores, MCP servers, and agent coordination tools are engineered to operate 100% locally with zero unconsented telemetry, analytics, or cloud data egress.
  • Unprivileged User Mode (Non-Elevation): Tools operate within standard user privileges and never require administrative elevation for normal agent, CLI, or GUI operation.
  • Integrity & Sandboxing: File operations, database transactions, process lifecycle calls, and hook integrations fail closed and adhere to strict local authorization and boundary checks.

There aren't any published security advisories