Skip to content

Security: ANFAIA/Agentic-Smart-Health

Security

SECURITY.md

Security Policy

Scope

This policy covers the Agentic Smart Health monorepo, including:

  • apps/agent-orchestrator — the multi-agent orchestration layer.
  • apps/slicer-mcp-server — the MCP server that bridges agents with 3D Slicer.
  • packages/core-schemas — shared Pydantic schemas for clinical data contracts.
  • packages/3dgs-engine — the 3D Gaussian Splatting rendering engine.

Given that this system processes sensitive clinical data (CBCT/DICOM, STL, clinical reports), security issues are treated with the highest priority.


Supported versions

Version Supported
main (pre-release) Yes
Tagged releases Yes (latest only)
Older releases No

Reporting a vulnerability

Do not open a public GitHub issue for security vulnerabilities.

This is especially critical for vulnerabilities that could:

  • Expose or leak patient clinical data (DICOM, STL, PDF reports, or any data handled by the ingestion agents).
  • Allow unauthorized access to or manipulation of the MCP server (slicer-mcp-server) or its exposed tools.
  • Compromise the integrity of the Digital Twin (unauthorized writes, data poisoning).
  • Bypass anonymisation or pseudonymisation mechanisms.
  • Violate GDPR, HIPAA, or equivalent clinical data protection regulations.

How to report

Please use one of the following private channels:

  1. GitHub private vulnerability reporting (preferred): navigate to the repository's Security tab and select Report a vulnerability. This keeps the report confidential until a fix is available.

  2. Email: send a detailed report to the maintainers at the address listed in the repository's GitHub profile. Encrypt your message with the maintainers' PGP key if available.

What to include in your report

  • A clear description of the vulnerability and its potential impact.
  • The affected component(s) and version(s).
  • Step-by-step reproduction instructions or a proof-of-concept (without exposing real patient data).
  • Any suggested mitigations or patches you may already have.

Response process

Step Target timeframe
Acknowledgement of receipt 48 hours
Initial severity assessment 5 business days
Fix developed and reviewed Depends on severity (critical: ≤ 14 days)
Advisory published (if applicable) Coordinated with reporter

We follow responsible disclosure: we will work with you to understand and fix the issue before any public disclosure. We will credit reporters in the release notes unless they prefer to remain anonymous.


General security guidelines for contributors

  • Never commit secrets, API keys, or credentials. Use .env files (excluded from version control via .gitignore) and reference .env.example for required variables.
  • Never commit real patient data to the repository, even in anonymised form. Use the synthetic datasets provided under data/.
  • Dependencies are pinned via uv.lock. Review dependency updates carefully; run make test after any lockfile change.
  • The MCP server exposes tools to external agent clients. Any new tool added to slicer-mcp-server must be reviewed for injection risks and excessive permission grants before merging.
  • Clinical data flowing through agents must remain within the authorised storage boundaries defined in the architecture (docs/architecture/). No agent may forward or persist clinical data to external services without explicit human approval.

Compliance references

This project is designed with the following regulatory frameworks in mind:

  • GDPR (EU General Data Protection Regulation)
  • HIPAA (US Health Insurance Portability and Accountability Act)
  • Applicable national health data protection regulations

Security decisions that affect regulatory compliance must be documented in docs/architecture/ and reviewed by a maintainer before merging.

There aren't any published security advisories