The Agent Skills Directory

Unverifiable Dependencies & Remote Code Execution (CRITICAL): The skill instructions include curl -s https://fluxcd.io/install.sh | sudo bash. This pattern of piping a remote script directly into a shell with root privileges is a major security vulnerability, as it allows for arbitrary code execution if the source or the connection is compromised.
Privilege Escalation (HIGH): The use of sudo in conjunction with a piped remote script unnecessarily grants root access to unverified external content.
External Downloads & Manifest Execution (HIGH): The command kubectl apply -f https://raw.githubusercontent.com/argoproj/argo-cd/stable/manifests/install.yaml executes remote configurations directly in the cluster. This bypasses security reviews and audit trails for infrastructure changes.
Data Exposure & Credential Access (HIGH): The skill provides commands to retrieve and decode administrative secrets (argocd-initial-admin-secret). While part of standard setup, exposing this in a script or skill context facilitates unauthorized credential access.
Indirect Prompt Injection (HIGH):
Ingestion points: External Git repositories defined in repoURL and Flux bootstrap commands.
Boundary markers: None. The agent is instructed to synchronize state directly from external repositories.
Capability inventory: The skill has the capability to create, modify, and delete any Kubernetes resource via argocd and flux controllers.
Sanitization: None. There is no validation of the manifests pulled from Git before they are applied to the cluster, allowing an attacker with repository access to gain full cluster control.

gitops-workflow