Stacker is a deployment tool that runs commands against user-owned servers and cloud accounts, generates production Docker Compose configurations, and manages secrets. We treat security seriously because the blast radius of a Stacker vulnerability can be a user's whole infrastructure.

Do not open a public GitHub issue for security vulnerabilities. Public disclosure before a fix is available exposes users to risk.

Report privately via one of these channels:

• Email: security@try.direct — preferred
• GitHub Security Advisories: Open a draft advisory

If you prefer to encrypt, request our PGP key in your first message and we'll respond with it before you send technical details.

To help us reproduce and prioritise, include as much of the following as you can:

• A clear description of the vulnerability and its potential impact
• Steps to reproduce (a proof-of-concept stacker.yml, command sequence, or minimal test case is ideal)
• The affected version (stacker --version) and, if relevant, target platform
• Whether the issue requires authentication, specific config, or specific timing
• Suggested remediation or mitigation if you have one

• Acknowledgement within 5 business days of receiving your report
• An assessment of severity and scope, shared back with you
• Regular updates (at least weekly) while we work on a fix
• Credit in the release notes for the fix, unless you prefer to remain anonymous
• Coordinated disclosure: we aim to publish a fix and a security advisory within 90 days of the initial report, and will discuss the exact timing with you

We are a small team and cannot promise a firm fix SLA, but critical issues are prioritised over feature work.

If you use an AI assistant to help draft your report, that is fine — please just verify the vulnerability actually reproduces before submitting. Reports that describe non-issues (a false positive from a scanner, a general threat pattern with no specific instance in Stacker) will be closed with a short note. Please do not submit speculative "AI CVE farming" reports — they consume volunteer time and slow down real disclosures.

Stacker is pre-1.0. Security fixes are released against the latest tagged release on the main branch. Users on older versions should upgrade to receive security fixes; we do not backport to previous minor versions.

The following are considered security issues we want to hear about:

• Stacker CLI binary — command injection, path traversal, deserialisation flaws, credential leakage in logs or generated files
• Casbin policies (access_control.conf) — privilege escalation, RBAC bypass, role confusion
• Server-side components (Status Panel agent, deployment engine) — remote code execution, unauthorised container control, secret exfiltration
• Generated deployment configuration — cases where Stacker generates a docker-compose.yml or Nginx config that exposes services more broadly than the stacker.yml intends (e.g. missing bind-to-localhost on database ports, missing reverse proxy on internal services)
• Vault / secrets integration — plaintext exposure, insecure defaults, cross-tenant leakage in the managed platform
• Template repositories (awesome-selfhosted-stacker) — templates that ship with weak defaults, insecure environment variables, or hardcoded credentials
• Supply chain — dependencies pulling in known-vulnerable code, malicious releases we've included

The following are not Stacker vulnerabilities and should be reported to the upstream project or handled by the operator:

• Vulnerabilities in third-party container images that templates deploy (report to the image maintainer — Ghost, Nextcloud, Postgres, etc.)
• Vulnerabilities in Docker, Docker Compose, Nginx Proxy Manager, Traefik, or the underlying Linux distribution
• Misconfiguration in user-provided stacker.yml (e.g. deliberately exposing a database port publicly)
• Attacks that require pre-existing root access to the target server, physical access, or social engineering of the user
• DoS via resource exhaustion on user-owned infrastructure (this is expected under adversarial load)
• Reports based solely on the output of a scanner without a demonstrated exploit path

If you are unsure whether something is in scope, err on the side of reporting. We would rather triage a false positive than miss a real issue.

Stacker's codebase includes a dedicated security test suite covering the categories most likely to introduce vulnerabilities in a deployment tool. As of this policy the suite includes:

• IDOR (Insecure Direct Object Reference) checks across deployments, projects, admin actions, ratings, and cloud resources
• Command injection tests against CLI-facing surfaces
• Authorization boundary tests against the agent, chat, server, and pipe interfaces
• SSH key handling and cloud credential isolation tests

See tests/security_*.rs for the current set. New surfaces that touch authorisation or secrets require a security test as part of the PR.

We publish a thanks section in release notes for reporters of valid vulnerabilities, unless the reporter prefers to remain anonymous. If you would like your GitHub handle, a personal site, or a company affiliation included, mention it in your report.

We do not currently run a paid bounty program. That may change as the project grows.

Available on request. Include a note in your initial email asking for our public key and we will respond before you send technical details.

Contact: security@try.directAdvisories: https://github.com/trydirect/stacker/security/advisoriesLicense: This project is MIT-licensed; this security policy applies to Stacker as distributed in this repository.