verification-loop
Audited by Socket on Sep 19, 2026
3 alerts found:
SecurityAnomalyx2SUSPICIOUS: the skill is internally coherent as a vulnerability-verification workflow, but its actual function is to give an AI agent offensive security capabilities with Bash, web access, evidence capture, and cross-skill orchestration. There is little sign of credential theft or malicious exfiltration, so this is not confirmed malware, but it is a high-risk offensive-security skill.
The fragment appears to be legitimate penetration-testing training and verification automation, not malware. It actively scans supplied targets and may perform SQL injection/XSS testing, so it must be restricted to authorized targets. The main security issue is unsafe interpolation of OUTDIR into shell and Python command strings, which can permit local command/code injection through a crafted output-directory argument. No credential theft, exfiltration, persistence, or destructive behavior is visible.
The fragment describes dual-use penetration-testing verification procedures and contains active exploit requests. It presents meaningful operational and authorization risks if executed against unapproved targets, and the automation lacks visible scope validation and robust input validation. However, the content is consistent with an explicitly stated authorized security-testing purpose and shows no clear supply-chain malware, credential theft, persistence, or covert exfiltration. Review and sandbox the scripts before use, enforce target allowlists, validate inputs, and require explicit authorization.