argos-edge is a security-adjacent project (reverse proxy + WAF + AppSec + SSO). Vulnerability reports are taken seriously; this document spells out what to report, where to send it, and what to expect back.
argos-edge is solo-maintained. Security fixes ship against:
| Version | Status |
|---|---|
Latest tagged release on main |
✅ Supported |
Previous minor (latest of 1.3.x if current is 1.3.y) |
✅ Supported on best-effort basis |
| Older releases | ❌ Not supported — upgrade to latest |
The latest release is always at
github.com/cmos486/argos-edge/releases.
Tooling-only patches (e.g. 1.3.36.X capture automation)
inherit the security posture of the panel binary they
target.
Do not open a public GitHub issue for security reports. Public reports give attackers a head start on un-patched deployments.
Preferred: GitHub Security Advisories private vulnerability reporting at github.com/cmos486/argos-edge/security/advisories/new. This routes the report directly to the maintainer in a private channel and tracks coordination metadata (severity, affected versions, patch ETA) in one place.
Alternative: open a GitHub issue with the title "Security report — coordination needed" and no technical detail. The maintainer will respond with a private channel.
In your report please include:
- Affected version(s) — output of
argos versionif available, or the docker image tag - Panel mode in use (
lan/behind_caddy) — relevant for surface-area scoping - Steps to reproduce, ideally against a fresh
docker compose up -dinstall - Impact assessment — what an attacker can do, and whether authentication or operator-mediated input is required
- Suggested fix or mitigation if you have one (optional)
This is a maintainer-time-permitting project. I aim to acknowledge reports within a reasonable timeframe and triage based on severity, but cannot commit to specific response SLAs. Active exploitation in the wild will be prioritized.
For coordinated disclosure, the industry-standard 90-day window applies as a default ceiling, with adjustments based on severity and exploitation status. If the vulnerability is being actively exploited against argos-edge deployments, I may ship a fix and disclose sooner to protect users.
If you do not hear back after a couple of weeks, please assume the report did not reach me and re-send via a different channel (e.g. a GitHub issue with the no-technical-detail title from the previous section).
In scope (argos-edge codebase):
- Panel binary (Go backend + embedded React SPA)
- Docker compose stack as published in this repo
- CLI subcommands (
argos backup,argos restore,argos reset-password, etc.) - HTTP API (
/api/...) authentication and authorization - Caddyfile bootstrap, CrowdSec acquisition + profiles config, AppSec rules — as committed in this repo
- Credential / secret handling
(
ARGOS_MASTER_KEY/ AES-GCM-encrypted DB columns) - TOTP enrollment + session lifecycle
- ForwardAuth path, OIDC sign-in flow
- WAF / AppSec rule bypass via panel-side input validation gaps
- Drift detection / reconciler false negatives that could mask a real bypass
Out of scope:
- Upstream Caddy 2 vulnerabilities — report to the Caddy project
- Upstream CrowdSec vulnerabilities (LAPI, AppSec component, bouncer plugin) — report to the CrowdSec project
- Upstream Coraza / OWASP CRS rule-set issues — report to Coraza / OWASP CRS
- Self-signed cert warnings during
lanmode HTTP-only panel access (intentional; documented indocs/getting-started/installation.md) - DoS by the operator against their own homelab stack
- Issues only reproducible on heavily-modified forks
- Theoretical attacks requiring privileged container access (if you already have root in the panel container, you have already won)
- Social-engineering attacks against the maintainer
If your report falls in the out-of-scope list above, the appropriate upstream project is the right destination — argos-edge cannot fix issues in code it does not own. The maintainer is happy to confirm whether a report is in scope before you spend effort writing it up; ping via the no-detail GitHub issue first if unsure.
See Security advisories for the public record.