verification-before-completion
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- [Prompt Injection] (SAFE): No override markers or jailbreak attempts were found. The instructions are focused on task verification and operational honesty.
- [Data Exposure & Exfiltration] (SAFE): No hardcoded credentials, sensitive file path access (e.g., .ssh, .aws), or network exfiltration patterns were detected.
- [Unverifiable Dependencies] (SAFE): The skill does not reference or download any external packages, scripts, or remote binaries.
- [Privilege Escalation] (SAFE): No commands for privilege escalation (e.g., sudo, chmod) or system configuration modifications are present.
- [Indirect Prompt Injection] (LOW): The skill involves processing output from external commands (tests, builds). This is a standard ingestion point where untrusted data could theoretically enter the context, but the skill implements this for the primary purpose of verification and includes warnings against trusting unverified reports.
Audit Metadata