Skip to content

Security: Cassandra-Collective/Ephesus

SECURITY.md

Security Policy

Supported versions

Ephesus is pre-alpha. There are no releases and no supported versions yet — the only thing that exists is the main branch. Security reports are still very welcome and will be taken seriously.

Version Supported
main Yes — report anything you find
Any release None exist yet

Reporting a vulnerability

Do not open a public issue for a security problem.

Report privately through GitHub's private vulnerability reporting, or by email to sensoymertefe@gmail.com with SECURITY in the subject.

Please include what you found, how to reproduce it, and what you think the impact is. A proof of concept helps enormously.

You can expect an acknowledgement within a few days. Since this is a solo-maintained pre-alpha project, please be patient with fix timelines — but you will get a real answer, not silence.

What is in scope

Ephesus runs real terminal agent CLIs as child processes on your machine, holds credentials for them, and writes to a coordination directory on disk. The interesting attack surface follows from that:

  • Renderer escape — anything that gets Node or filesystem access into the renderer process, or defeats contextIsolation / sandbox.
  • IPC validation gaps — a typed IPC handler in the main process that accepts input it should have rejected.
  • Secret leakage — a secret value reachable over IPC, written into logs, fixtures, the Agora, or an agent's transcript. Secrets are supposed to be write-only and reach agents only as declared environment variables.
  • Prompt injection through watched sources. The Stoa reads external repositories the user registers. That content is data, never instructions. A path where text from a watched source reaches code, prompts, or config without passing through a gated, human-approved proposal is a vulnerability, not a bug.
  • Agent boundary escape — an agent writing outside its own agora/agents/<id>/ directory, or committing to the Agora directly.
  • Gate bypass — anything that lets work proceed past a human approval gate, a budget, or the circuit breaker without the human.

What is out of scope

  • Vulnerabilities in the third-party agent CLIs themselves (claude, codex, gemini, …) — report those to their vendors.
  • Anything requiring an attacker who already has local code execution as your user. Ephesus is a desktop app that deliberately runs commands on your behalf; an attacker at that level has already won.
  • The absence of signing on builds. There are no builds yet; signed packaging is a planned milestone.

Disclosure

Please give a reasonable window to ship a fix before publishing. Credit will be given in the advisory unless you would rather stay anonymous.

There aren't any published security advisories