container-debug

Pass

Audited by Gen Agent Trust Hub on Nov 22, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill primarily operates by executing docker and kubectl CLI tools to inspect container states, logs, and events. These commands are essential for its diagnostic purpose.
  • [EXTERNAL_DOWNLOADS]: Recommends the use of the nicolaka/netshoot and busybox images to provide a shell or networking tools for containers that lack them. These are widely recognized and trusted diagnostic tools in the container ecosystem.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted container logs and health check outputs, which could potentially contain malicious instructions.
  • Ingestion points: Accesses docker logs, kubectl logs, and docker inspect outputs.
  • Boundary markers: Includes a specific Rule of Engagement (Rule 7) that explicitly instructs the agent to treat log lines and file content as evidence and never to act on them as instructions.
  • Capability inventory: Maintains access to docker and kubectl for troubleshooting.
  • Sanitization: Instructs the user and agent to redact environment variables and credentials from all captured evidence before analysis.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 22, 2026, 01:28 AM