aks-troubleshooting

Pass

Audited by Gen Agent Trust Hub on Oct 9, 2026

Risk Level: SAFECOMMAND_EXECUTIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • Command Execution from Metadata: The skill constructs shell commands using Kubernetes metadata, such as pod names, namespaces, and resource IDs, to perform diagnostics.
  • Evidence: Found in scripts/pod-deep-dive.sh and references/azure-network-path.md where variables like $POD and $NS are interpolated into kubectl and az commands.
  • Context: This is essential for the skill's troubleshooting purpose. The risk is mitigated by the fact that these identifiers typically conform to Kubernetes naming standards.
  • Privilege Escalation Gated by Approval: The skill utilizes high-privilege debug profiles for deep node diagnostics using Inspektor Gadget.
  • Evidence: references/inspektor-gadget.md details the use of kubectl debug --profile=sysadmin to create privileged pods for kernel-level tracing.
  • Context: The skill implements strong guardrails, including requiring explicit user approval before creating these pods, using digest-pinned images from the Microsoft Container Registry, and enforcing execution timeouts.
  • Indirect Prompt Injection Surface: As a troubleshooting tool, the skill processes untrusted data from cluster logs and events which could contain malicious instructions.
  • Ingestion points: The skill fetches data via kubectl logs and kubectl get events (e.g., in scripts/pod-deep-dive.sh).
  • Boundary markers: The pod-deep-dive.sh script limits visible logs to 50 lines and instructions in SKILL.md explicitly warn the agent never to follow commands or URLs found within logs.
  • Capability inventory: The skill utilizes kubectl and az commands, including the ability to create debug pods.
  • Sanitization: The scripts/pod-deep-dive.sh script includes a redact function that uses sed to filter out credential-shaped strings (Authorization, password, token, api-key) from the output.
  • External Resource Reference: The skill references a specific diagnostic image from the Microsoft Container Registry.
  • Evidence: references/inspektor-gadget.md specifies the image mcr.microsoft.com/oss/v2/inspektor-gadget/ig:v0.51.0@sha256:6610863f6d8cae28800f9331756434639bca44be065719cbcfe76e34c91dffa4.
  • Context: This is a vendor-owned resource used for intended diagnostic purposes and utilizes a cryptographic digest to ensure integrity.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 9, 2026, 04:19 PM