nvshmem-troubleshoot-and-report-bugs
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill has a vulnerability surface for indirect prompt injection by processing external log files.
- Ingestion points: The skill processes raw logs, stdout/stderr, and command stanzas provided by users for troubleshooting, as defined in
SKILL.md. - Boundary markers: A mandatory "Human-in-the-Loop Execution Boundary" is established in
SKILL.md, requiring the agent to display all proposed commands and obtain explicit user approval before execution. - Capability inventory: The skill uses a limited set of read-only tools for artifact inspection, including
rg,sed,head,tail,wc,stat, andfile. - Sanitization: The skill relies on user-directed redaction of sensitive data (credentials, tokens, IPs) before bug report submission as per
references/bug-report-template.md, but lacks automated log sanitization during the diagnosis phase.
Audit Metadata