Skip to content

Security: JoseVelazcoH/system-one-sql-agent

Security

docs/SECURITY.md

Security Policy

Supported versions

Only the latest commit on main receives security fixes.

Reporting a vulnerability

Please do not open a public issue for security problems. Email velazco.joseh@gmail.com with:

  1. A description of the issue and its impact.
  2. Steps to reproduce it, including the question you asked if relevant.
  3. The commit you tested (git rev-parse --short HEAD).

You will get an answer within 7 days. Once a fix is available, the issue can be disclosed publicly with credit to the reporter, if desired.

Scope

The agent is meant to run locally, on 127.0.0.1. Keep that in mind when reporting:

  • In scope: any way to write to or modify a database (queries run through readOnlyQuery in src/db.ts, inside a READ ONLY transaction), a question that makes the agent leak API keys or the database password, reading data outside the configured databases, and starting paid LLM runs from outside localhost with the default settings.
  • Out of scope: attacks that require exposing the UI with HOST=0.0.0.0 without authentication, which is not a supported setup, and vulnerabilities in third-party services (LLM providers, Vercel AI Gateway, PostgreSQL). Report those to their owners.

Handling secrets

  • API keys and the database password live only in .env, which is gitignored. config.yaml stores the names of those variables, never their values.
  • Use a PostgreSQL role with read-only permissions. The READ ONLY transaction is a second line of defense, not a replacement for database permissions.
  • If you commit a secret by mistake, rotate it with the provider right away. Removing it from history is not enough.

There aren't any published security advisories