Skip to content

Security: Rickaym/seti

SECURITY.md

Security policy

Seti runs Claude Code agents with shell access on your machine, so a flaw in it can hand that access to someone else. Reports are welcome and handled privately.

Reporting a vulnerability

Report it privately through GitHub: open this repository's Security tab and choose Report a vulnerability. Do not open a public issue or pull request for it.

Include what an attacker can reach, the steps to reproduce, the Seti commit you tested, and how Seti was exposed (loopback only, a reverse proxy, or Cloudflare Tunnel and Access).

You get an acknowledgement within 3 days and an assessment within 7. A confirmed issue is fixed on main and credited in the advisory unless you ask otherwise. Please keep it private until the fix is released.

Supported versions

Seti is released from main; fixes land there and are not backported. Run a recent main.

Scope

In scope: anything that lets someone without the owner's sign-in drive Seti's API, read a secret or a connection value, escape an agent's sandbox or egress allowlist, act beyond an agent's Seti control level, change security settings from the EA or an agent, or break the audit chain. Public surfaces (share links, webhooks, the demo) are in scope too.

Out of scope: attacks that need the owner's own sign-in or shell on the host, an agent the owner gave Bypass or an unsandboxed setup, and denial of service from a client the owner admitted.

The SETI Protocol specifies the security model these rest on, and docs/security.md covers running Seti securely.

There aren't any published security advisories