The Agent Skills Directory

[EXTERNAL_DOWNLOADS / REMOTE_CODE_EXECUTION] (HIGH): The installation instructions command the agent to execute a remote manifest directly into the cluster: kubectl apply -n argocd -f https://raw.githubusercontent.com/argoproj/argo-cd/stable/manifests/install.yaml.
Evidence: The URL belongs to 'argoproj'. While a legitimate project, 'argoproj' is not on the Trusted GitHub Organizations list. Piped or direct application of remote content to a cluster is a high-risk pattern.
[CREDENTIALS_UNSAFE] (MEDIUM): The skill provides commands that explicitly extract and display sensitive administrative credentials in the shell/agent context.
Evidence: kubectl -n argocd get secret argocd-initial-admin-secret -o jsonpath="{.data.password}" | base64 -d in SKILL.md. This exposes the initial admin password directly to the agent's output/logs.
[INDIRECT PROMPT INJECTION] (HIGH): The skill creates a high-privilege bridge between external, untrusted Git repositories and the internal Kubernetes cluster state.
Ingestion points: repoURL fields in Application and ApplicationSet resources target external repositories (e.g., https://github.com/org/myapp-manifests.git).
Boundary markers: Absent. There is no instruction to the agent to validate the source or contents of the repositories before configuration.
Capability inventory: High. The skill utilizes kubectl and argocd CLI to modify cluster state, create namespaces, and deploy applications.
Sanitization: Absent. The agent is instructed to automate deployments directly from the specified paths.
[COMMAND_EXECUTION] (MEDIUM): The skill requires the agent to run powerful CLI tools (kubectl, argocd) with cluster-admin level impact, including port-forwarding and credential management.

argocd-gitops