Skip to content

Security: cboxdk/platform

Security

SECURITY.md

Security Policy

cboxdk/platform compiles application intent into Kubernetes objects. It performs no IO of its own, but what it emits — RBAC, Secrets, network exposure, pod security context — is applied to real clusters by its consumers, so a flaw here becomes a misconfiguration there.

Reporting a vulnerability

Do not open a public issue. Report privately via GitHub Private Vulnerability Reporting (repository → SecurityReport a vulnerability). This is a best-effort open-source project with no funded security team and no response-time guarantee; we'll respond as promptly as we can and coordinate disclosure with you. Good-faith research under this policy is authorized (safe harbor).

The reports we most want to see:

  • Secret leakage into a compiled object — a value passed as envSecret that ends up readable from the cluster (inlined in a pod spec, a ConfigMap, a label or an annotation) instead of compiled into a Secret.
  • Secret leakage into a plan — any path by which secret material reaches a Plan, which consumers render into browsers, return from APIs and write to logs. planAgainst() redacts Secret values by design; a way around that is a vulnerability, not a bug report.
  • Privilege escalation in emitted objects — a spec that produces a container with more capability than the intent asked for, or an RBAC rule wider than the namespace it was meant for.
  • Cross-tenant reachability — a spec for one namespace that compiles a reference reaching another.
  • Non-determinism — the same spec compiling to different bytes. That is the property plan/diff and drift detection rest on; losing it is a correctness bug with security consequences.

Scope

In scope: everything under src/.

Out of scope: what a consumer does with the output. This package does not apply manifests, hold credentials, authenticate, or talk to a cluster — the apply path, admission policy, and credential handling belong to the consumer (Cbox Cortex, Cbox Local, or your own). A finding that requires the consumer to have already applied attacker-controlled intent is a finding about the consumer.

Supported versions

The package is pre-1.0. Security fixes target the latest 0.x release only; older minors are not backported.

There aren't any published security advisories