Skip to content

Security: Hyper-AI-Lab/cyber-ai-team

Security

SECURITY.md

Security Policy

Cyber AI Team is designed for owner-visible autonomy and auditability, but it is still evolving. Please treat it like sensitive infrastructure when deploying it.

Reporting Vulnerabilities

Please report security issues privately through GitHub Security Advisories for this repository when available, or by opening a minimal issue that does not disclose exploit details and asking for a private contact path.

Do not include live credentials, access tokens, private URLs, customer data, or exploit payloads in public issues.

Supported Version

The supported version is the latest commit on main.

Security Expectations

  • Replace all default secrets before non-local deployment.
  • Prefer OWNER_PASSWORD_HASH over plaintext owner passwords.
  • Restrict CORS to the exact owner-console origin in production.
  • Keep staging/production external side effects approval-gated.
  • Keep backups fresh before irreversible operations.
  • Treat ERPNext as canonical business state and protect its admin credentials.
  • Do not expose Docker, PostgreSQL, Redis, Qdrant, Temporal, or internal service ports publicly.
  • Run the quality/release gates before deployment.

Host And Cloud Isolation

Cyber AI Team does not require the Google Cloud CLI, Application Default Credentials, a Google billing account, or a Google Cloud project. The deployed containers must not mount host home directories, gcloud configuration, gsutil credential stores, browser profiles, or cloud service-account keys.

Interactive cloud login, automatic ADC creation, automatic API enablement, and automatic billing-account linking are prohibited in application code, scripts, Compose configuration, and CI. This boundary is enforced by:

python3 scripts/gcp-isolation-check.py

If a future research or integration requirement genuinely needs GCP, provision it as a separate lab with a non-personal identity, dedicated project, separate billing account, explicit budget and quota limits, and isolated credentials. Do not expose those credentials to Cyber AI Team or to an autonomous coding agent with unrestricted host and browser access.

Secret Handling

The repository includes a high-confidence secret scanner at scripts/secret-scan.py. Run it before commits:

python3 scripts/secret-scan.py
python3 scripts/gcp-isolation-check.py

If a secret is accidentally committed, rotate it immediately. Removing it from the latest commit is not enough once the commit has been pushed.

There aren't any published security advisories