Skip to content

Security: 7wdn/7shell

Security

SECURITY.md

Security policy

7Shell handles credentials, private keys and live access to servers. Security reports are taken seriously and answered as quickly as they can be.

Reporting a vulnerability

Open a private security advisory on the project's repository (Security → Report a vulnerability). Please do not open a public issue for a suspected vulnerability.

Include what you did, what happened, what you expected, and the version or commit you tested. A proof of concept helps enormously. If you are not sure whether something counts, report it anyway — "it probably does not matter" is how most of these get missed.

What you can expect: an acknowledgement, an assessment of severity and reach, and a fix or a reasoned explanation. Credit in the release notes if you would like it.

Supported versions

No version has been released yet, so nothing is supported for production use. Once releases begin, fixes will target the latest minor release.

The security model in one screen

The full model, including the threat model and how each rule is tested, is in docs/security-model.md. The invariants that cannot be traded away for convenience:

# Invariant
1 A secret never leaves the process: passwords and key material only ever travel in.
2 No credential is sent to a server before its host key has been approved.
3 An unknown host key is never accepted automatically, and a changed host key is a hard stop.
4 A decryption failure is indistinguishable from any other decryption failure.
5 The vault is written atomically; an interrupted write leaves the previous file usable.
6 The interface never names a local path; it receives tokens for files the user picked.
7 Every privileged or destructive action has a policy decision and an audit event.

What is in scope

The desktop application, the Rust crates, the descriptors, the build and release tooling, and the supply chain of the dependencies themselves. Out of scope: attacks that require an already compromised machine or a malicious server you deliberately connected to — although a server that can make the client leak something is in scope, and is interesting.

There aren't any published security advisories