The Agent Skills Directory

[PROMPT_INJECTION] (LOW): Potential surface for indirect prompt injection in scripts/debug_pod.py and SKILL.md debugging workflows. Malicious instructions could be embedded in Kubernetes pod logs or event messages to influence the agent's subsequent actions.
Ingestion points: Untrusted data enters the agent context via kubectl logs and kubectl get events in scripts/debug_pod.py.
Boundary markers: Log output is wrapped in markdown code blocks (```) for delimitation.
Capability inventory: The skill allows high-privilege operations including kubectl exec, kubectl delete, kubectl patch, and kubectl port-forward.
Sanitization: No specific sanitization or filtering of log content is performed prior to display.
[COMMAND_EXECUTION] (SAFE): The skill performs administrative command execution via kubectl. This is the primary intended purpose of the skill. Destructive commands are mitigated by strong instructional guardrails requiring context verification and dry-run flags.

kubernetes-operations