The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute multiple shell commands to manage infrastructure state.
The agent is prompted to use kubectl for querying and inspecting cluster resources, such as applications, CRDs, and secrets as seen in Workflows/DiagnoseSyncFailure.md and Workflows/OnboardCluster.md.
Git commands including git add, git commit, and git push are used to synchronize local changes with remote infrastructure repositories in Workflows/CreateRootApplication.md and Workflows/OnboardCluster.md.
Supporting tools like kustomize, jq, and the argocd CLI are utilized for manifest processing and validation.
[INDIRECT_PROMPT_INJECTION]: The skill demonstrates an attack surface for indirect prompt injection by ingesting and acting upon data from the execution environment.
Ingestion points: The agent reads output from kubectl get application -o yaml in Workflows/DiagnoseSyncFailure.md and directory listings via ls or grep in Workflows/CreateRootApplication.md.
Boundary markers: The instructions lack explicit markers or delimiters to separate tool output from instructions, creating a risk of accidental obedience if external data is maliciously crafted.
Capability inventory: Across the workflow scripts, the agent has capabilities to perform file writes, modify Git repositories (git push), and update cluster state (kubectl apply, argocd app create).
Sanitization: There is no evidence of sanitization, escaping, or validation of the content received from external command outputs before it influences agent logic.

ArgocdClusterBootstrapping