The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill's documentation and example files (e.g., SKILL.md, examples/core.md) contain placeholder credentials and sensitive strings within manifest examples, such as sk-live-abc123 and postgresql://user:pass@db:5432/app. These are clearly marked for educational purposes.
[EXTERNAL_DOWNLOADS]: The Helm chart examples in examples/helm.md reference the Bitnami repository (https://charts.bitnami.com/bitnami), which is a well-known and reputable source for official Kubernetes packages. Additionally, metadata.yaml references a schema from the author's official repository.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion and processing of Kubernetes manifests and Helm charts, creating a surface for indirect prompt injection.
Ingestion points: User-provided or external YAML manifests and Helm charts (as seen in SKILL.md and examples/core.md).
Boundary markers: While the skill provides extensive security requirements for the manifests themselves, it lacks specific instructions or delimiters to prevent the agent from following instructions potentially embedded in comments or metadata within the ingested data.
Capability inventory: The reference.md file documents and encourages the use of powerful CLI tools such as kubectl (e.g., apply, exec, port-forward) and helm for cluster interaction.
Sanitization: There are no documented procedures for sanitizing or validating the content of these manifests before they are processed or applied using the mentioned tools.

infra-containers-kubernetes