The Agent Skills Directory

[COMMAND_EXECUTION]: The skill references several shell scripts in its documentation (scripts/argo_flux_install.sh, scripts/kustomize_apply.sh, scripts/git_push.sh) designed to automate Kubernetes installation and deployment tasks. While these scripts are described as bundled resources, executing scripts that interact with Kubernetes clusters and Git repositories is a high-privilege activity that should be monitored.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest user-provided requirements to generate infrastructure manifests and automation code, creating an attack surface for indirect prompt injection.
- Ingestion points: User requests detailing GitOps setups, target clusters, and application requirements (described in SKILL.md).
- Boundary markers: No specific delimiters or "ignore instructions" warnings are mentioned to separate user-provided data from the agent's generation logic.
- Capability inventory: The skill can generate ArgoCD Application manifests, Flux Kustomization files, and execute referenced shell scripts to push changes or apply configurations to Kubernetes.
- Sanitization: The documentation does not describe any sanitization or validation processes for user-supplied data before it is incorporated into generated YAML or shell code.

building-gitops-workflows