The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on extensive use of kubectl commands to perform cluster diagnostics and management, including kubectl exec for running arbitrary commands inside containers and kubectl delete for removing resources.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and display untrusted data from external sources, creating a potential attack surface for indirect prompt injection.
Ingestion points: Commands such as kubectl logs, kubectl get events, and kubectl describe pod read data from container outputs and cluster events in SKILL.md and references/error-decoder.md.
Boundary markers: There are no explicit instructions or delimiters used to warn the agent to ignore potentially malicious instructions embedded in logs or event messages.
Capability inventory: The agent has significant capabilities including executing shell commands (kubectl exec), modifying cluster state (kubectl apply, kubectl delete), and transferring files (kubectl cp) as documented in references/kubectl-cheatsheet.md.
Sanitization: No sanitization or validation of the ingested log/event content is performed before processing.
[CREDENTIALS_UNSAFE]: The skill provides instructions for managing and viewing sensitive information, which could lead to accidental exposure of credentials.
Evidence: SKILL.md includes commands for creating secrets with plaintext passwords (kubectl create secret docker-registry --docker-password=<password>) and references/kubectl-cheatsheet.md provides commands to decode and view secret data in the terminal (kubectl get secret <name> -o jsonpath='{.data.password}' | base64 -d).
[EXTERNAL_DOWNLOADS]: The skill suggests running debugging tools from external container images.
Evidence: In references/kubectl-cheatsheet.md, the skill recommends using the nicolaka/netshoot image for networking debugging and the busybox image for general debugging, which are external images pulled from public registries.

kubernetes-troubleshooting