This project has two kinds of "security" — software security in the conventional sense, and patient safety in the medical-information sense. We treat both as the highest priority.
If you believe content in this repository could lead patients to harm
themselves — e.g. a compound listed without proper contraindications, a
hypothesis written in a way that reads as clinical recommendation, a
trigger framing that could push patients into unsafe self-experimentation —
open an issue tagged safety immediately.
We will:
- Acknowledge within 24 hours.
- Hide the offending content (issue / PR / file section) within 48 hours if the risk is real and ongoing.
- Publish a public correction with rationale.
Examples of past / hypothetical safety reports:
- "The maintenance.md doc mentions LDN but doesn't flag that it requires prescription + clinician supervision."
- "The trigger doc lists aspirin without flagging that NSAID-intolerant MCAS patients can have severe reactions."
If you discover a vulnerability in any code in this repository (we run no production services; this is essentially client-side cheminformatics software), do not open a public issue. Instead:
- Open a private GitHub Security Advisory:
Securitytab →Report a vulnerability. - For private channels outside GitHub, see CONTACT.md. The responsible entity is MR Dula Medical (DBA of MR Dula Enterprise, LLC, Raleigh, NC, USA).
We will respond within a week. Critical fixes ship same-day where possible.
- Code in
scripts/,notebooks/,.github/workflows/. - Supply-chain risk (deps in
pyproject.toml,requirements-colab.txt). - Data integrity in
data/(SMILES injection, malformed CSVs leading to unsafe downstream processing).
- The accuracy of any prediction in
outputs/. Predictions are hypotheses, not security claims. Use the hypothesis_proposal template for those. - The robustness of any third-party service we reference (PubChem, PyTDC, Colab, Enamine, AlphaFold). Report those upstream.
For software vulnerabilities, we follow standard 90-day coordinated disclosure: report → acknowledge → patch → publish. We will credit reporters who request it.
- Personally identifying patient information.
- Credentials, keys, or tokens.
- Payment.
- We do not collect telemetry or analytics.
- The repo has no backend; running scripts/notebooks does not phone home.
- The only network calls are to public scientific APIs (PubChem, AlphaFold, PyTDC) and only when explicitly invoked.