Skip to content

Security: prod-explore/everything-app

SECURITY.md

Security policy

Supported versions

Only main. There are no tagged releases; fixes land on main and nothing is backported.

Reporting a vulnerability

Please do not open a public issue. Use GitHub's private vulnerability reporting instead: Security → Report a vulnerability on https://github.com/prod-explore/everything-app.

Include what you found, how to reproduce it (a request, a command, a compose file) and the commit you tested. This is a one-person project: reports are handled on a best-effort basis and normally acknowledged within a week.

Scope

The project's security claims are listed in the README. A report is in scope when it breaks one of them:

  • Sandbox escape. Code running in a sandbox reaches the host, the Docker socket, the supervisor, another sandbox, another service's network or the app's database.
  • Key exposure. A provider API key, KEY_VAULT_SECRET, a service token, a session cookie or a GitHub credential reaches a sandbox, the browser agent, the model, a log, an API response or the page.
  • SSRF. web_fetch or the browser agent can be made to reach a private, link-local or metadata address, or a URL the model composed itself is fetched without the approval it needs.
  • Authentication. Reaching the HTTP API, /internal/*, the supervisor, sandbox-mcp or playwright-mcp without the required password, session or token.
  • Approval gate. A tool call with a side effect, or a push to a protected branch, that runs without the approval the policy asks for.

Out of scope: an operator who deliberately weakens the setup (ALLOW_WEAK_SECRETS, ports published beyond 127.0.0.1, an egress guard that is not installed), and what a model does inside the limits a person approved. Advisories in third-party packages are welcome when they are reachable from this code; npm audit already runs in CI.

There aren't any published security advisories