fhirmini is a development-sandbox FHIR platform. Out of the box it is not configured for, and must not be used with, real Protected Health Information (PHI) or any production patient data. Running real PHI is a legal and operational process — see the PHI section in the README. The default posture (PHASE=dev-sandbox) and the AI layer's cloud-backend guard exist to keep you on the synthetic-data path.

Do not open issues, pull requests, logs, or test fixtures that contain real patient data. If you believe a contribution or artifact contains real PHI, report it privately (below) — do not comment publicly.

Please report security vulnerabilities privately, not via public issues:

• Use GitHub's "Report a vulnerability" (Security Advisories) on this repository, or
• Email the maintainers (see the repository's profile / CODENAME contact).

Include: affected component (FHIR core / engine / AI sidecar), version or commit, reproduction steps, and impact. We aim to acknowledge within a few business days.

• The FHIR server and all services bind to loopback (127.0.0.1) by default; external exposure is intended only via an authenticated Cloudflare Tunnel + Access (see scripts).
• Secrets live in .env (gitignored). Never commit credentials.
• The integration engine disables referential-integrity-on-write by design (messages can arrive out of order) — treat ingested data as untrusted and validate downstream.
• The AI local backend keeps data on-device; the openrouter backend sends content to a third party and is hard-disabled when PHASE=phi-readiness.

This is pre-1.0 software; security fixes are applied to the main branch.