network-troubleshooting
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill implements robust safety boundaries that explicitly prohibit the extraction or printing of credentials, tokens, proxy URLs, or sensitive configuration data.\n- [SAFE]: Instructions strictly forbid bypassing TLS/SSL certificate verification (e.g., using
--insecureorNODE_TLS_REJECT_UNAUTHORIZED=0), ensuring diagnostics do not compromise security.\n- [SAFE]: The diagnostic workflow is read-only and target-scoped, requiring explicit user consent before any remediation actions are taken, preventing unauthorized system changes.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted error data, but limits risks via strict instructions and target scoping.\n - Ingestion points: User-provided error text and target hostnames.\n
- Boundary markers: Explicit Safety Boundaries section.\n
- Capability inventory: Read-only diagnostics (ping, curl, nslookup, openssl).\n
- Sanitization: Instructions require replacing internal hosts with placeholders.
Audit Metadata