Skip to content

Security: AndreasEhstandLicenseofClarityLOC/augmanitai

Security

SECURITY.md

Security Policy

Scope

AUGMANITAI is primarily a terminology database, but the accompanying tooling (Python package, MCP server, protocol adapters, JSON-LD / SHACL / TBX generators and validators) can process untrusted input and therefore has a security surface.

In scope: the MCP server (protocols/mcp/), protocol bridges (protocols/bridges/), ontology/provenance generators and validators, and any CI/CD configuration.

Out of scope: third-party dependencies (report upstream), clearly-marked non-production demo code, theoretical attacks without a practical exploit, social engineering, and the editorial content of the terminology definitions themselves (use normal Issues for those).

Reporting a Vulnerability — responsible disclosure

Please report security issues privately. Do not open a public GitHub issue for a security problem.

  • Preferred channel: private GitHub Security Advisory for this repository.
  • Fallback channel: email ehstand.schule@gmail.com with the subject line AUGMANITAI SECURITY. For sensitive details, send a first non-sensitive message and request a PGP key fingerprint before transmitting the full report.

Please include: a clear description, steps to reproduce (PoC if available), affected version(s), your impact assessment, and any suggested mitigation.

Response timeline (best effort, single-maintainer project)

  • Acknowledgment: within 72 hours.
  • Triage & severity assessment: within 7 days.
  • Coordinated disclosure plan: within 14 days.
  • Patch: depending on severity, typically within 30 days for high-severity issues.

Safe Harbor

This project supports good-faith security research. If you make a good-faith effort to comply with this policy, we will consider your research authorized, work with you to resolve the issue, and will not pursue legal action related to that research.

Disclosure

After a fix is released we will publish an advisory, credit the reporter (unless anonymity is requested), and request a CVE if warranted.

Contact


Bound by the Ethical Disclaimer §1–§20 (DISCLAIMER.md).

There aren't any published security advisories