The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The runbook includes a command to extract and decode a Kubernetes secret (vastai-secrets). The command kubectl get secret vastai-secrets -o jsonpath='{.data.api-key}' | base64 -d directly exposes sensitive API keys in plaintext within the agent's execution environment.
[COMMAND_EXECUTION]: The skill requires and utilizes high-privilege access to a production Kubernetes cluster via kubectl. It performs sensitive operations including secret creation (kubectl create secret), environment variable modification (kubectl set env), and deployment restarts (kubectl rollout restart).
[COMMAND_EXECUTION]: The skill attempts to execute a script located at ./scripts/vastai-debug-bundle.sh. Since this script is not part of the provided skill files, its logic is unverifiable and presents a risk of executing unintended or malicious code present on the host system.
[COMMAND_EXECUTION]: The skill is susceptible to Indirect Prompt Injection. Evidence: (1) Ingestion points: kubectl logs and curl https://api.yourapp.com/health. (2) Boundary markers: None present. (3) Capability inventory: Full kubectl access and bash execution. (4) Sanitization: No escaping or validation is performed on log or endpoint data before processing.

vastai-incident-runbook