Skip to content

Security: stellar-experimental/stellar-raven

SECURITY.md

Security policy

This repository backs a live, internet-facing service: the Stellar Raven MCP gateway at raven.stellar.org (a Cloudflare Worker acting as an OAuth authorization server and a sandboxed-code executor). Security reports are very welcome.

Reporting a vulnerability

Please do not open a public issue for security problems.

  • Preferred: GitHub private vulnerability reporting on this repository (Security tab → "Report a vulnerability").
  • Alternatively: email frontier@stellar.org with a description, reproduction steps, and the impact you believe it has.

You can expect an acknowledgement within a few business days. Please give us a reasonable window to remediate before any public disclosure.

For anything that is not a vulnerability — connection problems, catalog questions, general support — use the #raven channel in the Stellar Developers Discord. Do not post a vulnerability there.

Scope

  • This repository's source and generated artifacts.
  • The deployed gateway at raven.stellar.org, including any retired hostname still routed to it (auth flows, the search/execute MCP surface, the /playground browser surface and its login/chat routes, sandbox isolation/egress).

Out of scope: the upstream services the gateway aggregates (Lumenloop, Stellar Light/Scout, Stellar Docs) — report those to their respective owners; vulnerabilities requiring a compromised maintainer machine; volumetric denial of service.

Notes for researchers

  • Model-authored code runs in a Dynamic Worker isolate with no network egress (globalOutbound: null); all service traffic goes through host-side adapters that hold the secrets. Sandbox-escape or egress findings are the most valuable class of report here.
  • Auth design (WorkOS OAuth + named API-key gate) is documented in ARCHITECTURE.md and research/auth-workos.md; README.md links to those operational details. The design does not depend on secrecy, and reports that only restate documented behavior (e.g. that a key bypass exists) are not vulnerabilities.

There aren't any published security advisories