debug-k8s-collection

Pass

Audited by Gen Agent Trust Hub on Jun 25, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill contains various commands for kubectl, helm, and python3. A critical security boundary is established by the 'Agent sandbox note' which explicitly forbids the AI from executing these commands itself. Instead, the AI is instructed to present these commands to the user for manual execution in their own terminal, ensuring human-in-the-loop oversight for all administrative actions.
  • [CREDENTIALS_UNSAFE]: Step 3 provides a Python command that decodes the agent-credentials Kubernetes secret to display its length and partial contents (first 8 and last 4 characters). While this interacts with sensitive data, it is designed as a diagnostic tool for the user to verify token integrity without exposing the full secret in logs or the terminal. This is a standard troubleshooting pattern for validating OTel ingest tokens.
  • [EXTERNAL_DOWNLOADS]: The skill suggests running a temporary pod using the curlimages/curl image for network reachability testing. This is a well-known, community-standard utility image used for network diagnostics and does not pose an inherent security risk in this context.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 25, 2026, 04:46 PM
Security Audit — agent-trust-hub — debug-k8s-collection