This is the public, synthetic-only workbench for proposing analyses that may eventually run beside private Brainstem data. It lets researchers and coding agents implement a paper, reproduce its public method, and prove that the container returns a bounded result without receiving participant data.
Nothing in this repository is approved for a private cohort. A passing test, pending proposal, or published container is not medical validation, a clinical claim, data access, or permission to release a result.
- Develop only against the generated fixtures in
synthetic/. - Never add participant records, production endpoints, credentials, wallets, device identifiers, or screenshots of health data.
- Cohort results must preserve the 20-participant disclosure floor and return only the bounded result contract.
- Containers must not require network access or write outside their output directory.
- Scientific, privacy, and security decisions are independent and bind the exact source digest.
- Production image scanning, Ocean publication, dataset allowlisting, study approval, and participant consent happen outside this repository.
Requirements are Python 3.11+ and Docker.
python3 -m unittest discover -s tests -v
python3 scripts/contribution_gate.py validateThe contribution guide contains the constrained container check and the path from a public paper to a reviewable candidate: CONTRIBUTING.md. Coding-agent users should also read contributions/AI_AUTHORING.md.
All included proposals and review records are public work in progress unless their exact source digest has three named approvals. The protected release workflow additionally requires a maintainer-controlled GitHub environment.
Licensed under GPL-3.0. Security reports belong in GitHub's private vulnerability reporting flow; see SECURITY.md.