The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation and reference guides provide instructions for installing essential cluster components from external sources.
Evidence:
Fetches the Kubernetes Metrics Server configuration from the official kubernetes-sigs repository on GitHub.
Downloads the OPA Gatekeeper deployment manifest from its official GitHub repository.
References the cert-manager installation manifest from the official project repository on GitHub.
Clones the official Kubernetes autoscaler repository for Vertical Pod Autoscaler (VPA) setup.
All identified targets are well-known technology projects or trusted organizations, and these downloads are documented neutrally as standard operational procedures.
[COMMAND_EXECUTION]: The skill includes several scripts designed to automate cluster maintenance and auditing tasks.
Evidence:
scripts/audit-networkpolicies.sh, scripts/cost-analysis.sh, and scripts/validate-resources.sh execute kubectl commands to retrieve cluster state and jq/awk for data processing.
Python scripts in examples/python/ utilize the official kubernetes client library to perform security and resource audits.
These operations are consistent with the skill's primary purpose of Kubernetes operations and troubleshooting.
[INDIRECT_PROMPT_INJECTION]: The auditing and reporting scripts ingest untrusted data from the Kubernetes cluster, creating a theoretical injection surface.
Evidence:
Ingestion points: Multiple scripts (e.g., scripts/cost-analysis.sh, examples/python/validate_rbac.py) read pod labels, annotations, and resource configurations from the cluster environment via kubectl and the Python SDK.
Boundary markers: There are no explicit delimiters or specific warnings used to isolate the agent from potentially malicious content embedded in cluster metadata during processing.
Capability inventory: The skill possesses capabilities for cluster interaction (read/audit) and shell execution, though it does not automatically perform write operations based on ingested metadata.
Sanitization: Data processing relies on standard JSON parsing through jq and the kubernetes Python library; no custom sanitization or filtering of input strings is implemented.

operating-kubernetes