k8s-aks-troubleshooting

Pass

Audited by Gen Agent Trust Hub on Nov 10, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted external data, but includes robust mitigations to ensure safety.
  • Ingestion points: The agent ingests data from kubectl logs, Kubernetes events, and Azure activity logs across all reference files to diagnose failures.
  • Boundary markers: The skill establishes a strict "Read first, ask before changing" policy. Write operations are clearly partitioned into "Ask first" (requiring explicit approval) and "Proposed fix" (intended for the user to run manually).
  • Capability inventory: The skill possesses the capability to modify cluster state via kubectl annotate, kubectl scale, kubectl delete, kubectl rollout restart, and Azure state via az aks update and az role assignment create (SKILL.md, references/load-balancers.md, references/nodepools-upgrades.md).
  • Sanitization: The instructions explicitly command the agent to never read Kubernetes Secrets and to only print the names of environment variables, rather than their values, significantly reducing the risk of accidental data exposure.
  • [DATA_EXFILTRATION]: The skill includes proactive safety instructions to prevent data exposure. It specifically mandates: "Never read Secrets, and do not print tokens, keys or client secrets from any source" (SKILL.md) and instructs the agent to only fetch environment variable names to avoid sensitive data leakage (references/storage-identity.md).
  • [COMMAND_EXECUTION]: All commands provided are standard kubectl and az CLI operations used for infrastructure diagnostics. The skill distinguishes between read-only investigations and state-changing operations, requiring a human-in-the-loop for any modifications.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 10, 2026, 11:25 PM