web2-recon

SUSPICIOUS/HIGH-RISK but not confirmed malware. The skill’s behavior matches its stated offensive recon purpose, yet that purpose itself gives an AI agent substantial security-testing capability against arbitrary targets and relies on a wide, partially unpinned third-party toolchain. Data flows mostly go to official target/services rather than exfiltration endpoints, so this looks more like a high-risk offensive workflow skill than credential-stealing malware.