The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill identifies and processes external data from Git repositories (YAML manifests, Helm charts, Kustomize overlays). This creates an attack surface where malicious instructions could be embedded in the managed repositories.
Ingestion points: Repo URLs and paths defined in the Application and AppProject specs.
Boundary markers: None explicitly mentioned for isolating manifest content from instructions.
Capability inventory: Executes CLI commands (argocd, kubectl) to synchronize state and retrieve logs.
Sanitization: The skill recommends protecting the main branch with PR checks to ensure only reviewed configurations are tracked.
[COMMAND_EXECUTION]: The instructions direct the agent to use the argocd CLI for operations like sync, diff, rollback, and get. These are expected tools for the stated purpose of managing Argo CD.
[EXTERNAL_DOWNLOADS]: The skill references official documentation and GitHub issues from the Argo CD project (argoproj). These are trusted, well-known technical resources.

argocd