The Agent Skills Directory

[SAFE]: The skill instructions explicitly state that the agent should never execute cluster commands like kubectl or helm, ensuring all sensitive operations remain under the user's control.
[SAFE]: The workflow includes a dedicated 'Intake' phase where the user is prompted to choose a sensitive-data handling preference (redact, mask, or share as-is) before providing any manifest content.
[SAFE]: No evidence of prompt injection, obfuscation, or unauthorized data exfiltration was found. The skill operates purely within the provided workspace and intended workflow.

antithesis-k8s-onboarding-assistance