Skip to content

Security: Kerr1st/Second-Brain

Security

SECURITY.md

Security Policy

Second Brain is a local-first, single-user AI memory system. It is not designed to be exposed as a hosted multi-tenant service.

Supported Use

  • Run the MCP server over stdio from a trusted local AI client.
  • Keep PostgreSQL bound to localhost.
  • Store real credentials in environment variables, a local secret manager, or user-local config files.
  • Encrypt backups before syncing them off-machine.

Do Not Commit

  • API keys, OAuth tokens, Gmail app passwords, or AWS credentials.
  • Database dumps, raw memory exports, logs, or personal notes.
  • Backup keys or decrypted backup material.
  • Browser cookies or captured session data, except bounded real-data test fixtures deliberately reviewed under tests/fixtures/ as described below.

Reviewed Real-Data Fixtures

Agent Task capture tests may commit bounded excerpts of real task history under tests/fixtures/. This deliberate exception is recorded in ADR 0005. Before committing such a fixture, review it for credentials, tokens, private keys, cookies, database exports, and unrelated personal material. Accepted fixture content and local path metadata become permanent Git history; deleting the file later does not erase it from existing clones or prior commits.

Use .env.example as the shape of local configuration. Copy it to .env and replace placeholders locally; .env is gitignored.

Reporting

This is a personal public proof artifact, not a production service. If you find a security issue in the public code, open a GitHub issue with reproduction details but do not include secrets, private memory content, or exploit payloads against third-party services.

Related Documentation

There aren't any published security advisories