failsafe-security/artex

What it is

FailSafe ARTEX is a source-built distribution of ARTEX, an open-source multi-agent security testing harness. The fork was created after the prebuilt tomicdd/artex Docker image was reported to contain an unauthenticated command-execution backdoor in November 2026. This build compiles the audited v0.3.15 source tag and preserves the upstream runtime stack: Node 22, Playwright, Chromium, and nmap.

The deployment differs from upstream in a few concrete ways: the console binds to 127.0.0.1:8787 instead of 0.0.0.0, the recording proxy is not published, no-new-privileges is added, the JWT key is persisted, base images are pinned by digest, and the self-update path points at this fork with one-click update disabled. The API, UI, skills, browser MCP, and browser-driven recon remain unchanged.

It is aimed at security testers who want to run the ARTEX agent stack locally without pulling the reported-backdoored upstream image.

Quick start

You need Docker Engine with Compose v2 and roughly 5 GB of free disk for the browser stack and Chromium. If only the standalone docker-compose command is available, substitute it for docker compose below.

cp .env.safe.example .env.safe
chmod 600 .env.safe
# edit .env.safe: set POSTGRES_PASSWORD and, optionally, an LLM key
docker compose --env-file .env.safe -f docker-compose.safe.yml up -d --build

Then open the local console and complete the first-run setup. If you left the LLM key blank, configure a provider in the console before starting a scan. Status and logs are checked with:

docker compose --env-file .env.safe -f docker-compose.safe.yml ps
docker compose --env-file .env.safe -f docker-compose.safe.yml logs -f artex

To stop without deleting data, run docker compose --env-file .env.safe -f docker-compose.safe.yml down; adding -v removes the database and ARTEX data as well.

There is also a bare-binary path: build the web UI into server/webui/dist, compile with CGO_ENABLED=0 go build -tags embedui, then run ./artex -addr 127.0.0.1:8787 -proxy 127.0.0.1:8788 with ARTEX_PG_DSN pointing at a PostgreSQL database you control. That path does not install Node, Playwright, Chromium, or the recon CLI tools.

Build verification

make verify checks for the known implant indicators (verifyCaptcha, tomicdd, congdonglinux), prints the binary SHA-256, and lists the image ID. The full audit of the tag is in the audit report.

Constraints and data handling

ARTEX is an autonomous agent that can execute shell commands and send network traffic; it should only be used against systems you own or have explicit written permission to test. The container runs as root, matching upstream, because raw nmap scans and runtime apt/npm -g installs depend on it. The hardening here is limited to loopback-only console binding, an unpublished proxy, no-new-privileges, a persisted JWT key, and digest-pinned base images.

.env.safe holds the database password and optional LLM keys; keep it at 0600 and out of version control. Scan evidence may include credentials, tokens, and customer data, so /app/data and the database should be treated as sensitive. Back up the pgdata, artex_data, and artex_key volumes together. Updates require reviewing the upstream diff, rebuilding, running make verify, and updating the audit report first, as described in the upstream notes. The project is AGPL-3.0, with modifications released under the same license.

HotRepo's take

The repository appeared on the new-project list on 2026-11-10 with 107 stars gained over a few hours, but it has no recorded GitHub daily rank. That single snapshot only establishes the day's counts; it does not show sustained growth. The most important thing for a reader to check is whether the source-built binary and the audit claims can be reproduced, since the project's central premise is that a different distribution was compromised. The README gives concrete verification commands and an audit report, but the trust argument rests on those artifacts rather than on the repository's age or star count.