Skip to content

Security: baokhang83/mnemo-warmup

SECURITY.md

Security Policy

Supported Versions

mnemo-warmup is currently in active pre-1.0 development (alpha). Security fixes are applied to the latest release line on main, and to 0.2.x until 0.3.0 is released as final.

Version Supported
0.3.x (beta) ✅
0.2.x ✅ until 0.3.0 final
< 0.2.0 ❌

Reporting a Vulnerability

Please do not report security vulnerabilities through public GitHub issues or discussions.

If you discover a security vulnerability in mnemo-warmup (including issues related to RBAC scope, credential handling, archive integrity verification, or container privilege escalation), please report it privately:

  1. GitHub Private Vulnerability Reporting (Preferred): Go to the Security tab of this repository and click "Report a vulnerability" to open a private security advisory.
  2. Response Timeline: You will receive an acknowledgment within 48 hours. Once confirmed, we will work on a patch and coordinate a disclosure timeline with you before publishing a fix.

Security Model & Scope

For details on mnemo-warmup's least-privilege Kubernetes RBAC, secret isolation, registry credential boundaries, and fail-safe execution guarantees, see docs/security.md.

Out of Scope

  • Application startup side effects: By design, the asynchronous trainer CronJob executes a real instance of your application image inside the cluster (spring-refresh or http-probe). Operators are responsible for reviewing mnemo-warmup inject --preview and ensuring their training environment isolates production databases, external APIs, or scheduled jobs.
  • Registry storage growth: mnemo-warmup intentionally omits registry delete permissions (ecr:BatchDeleteImage). Unbounded archive growth is managed by the operator's OCI registry lifecycle policy.

There aren't any published security advisories