The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external Kubernetes APIs, which may contain attacker-controlled content in metadata or status fields.
Ingestion points: The scripts/discover_apis.py script and the kubectl commands defined in references/operator-checks.md read state directly from the cluster.
Boundary markers: The diagnostic scripts produce structured JSON output, providing clear boundaries for the agent.
Capability inventory: The skill uses Bash to run kubectl and uv to execute Python scripts.
Sanitization: The scripts do not perform explicit validation or sanitization of string-based metadata (e.g., event logs or resource names) before inclusion in the diagnostic report.
[COMMAND_EXECUTION]: The skill invokes system commands to interact with the cluster and run diagnostic logic.
Evidence: SKILL.md provides instructions for running kubectl commands and executing scripts via uv run.
[EXTERNAL_DOWNLOADS]: The skill downloads the kubernetes Python library to facilitate cluster discovery.
Evidence: scripts/discover_apis.py specifies kubernetes>=28.1.0 as a dependency. This is an official and widely trusted package for Kubernetes integration.

kubernetes-health