jndi-injection
Warn
Audited by Gen Agent Trust Hub on Apr 9, 2026
Risk Level: MEDIUMDATA_EXFILTRATIONREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill documents specific JNDI lookup patterns intended to leak sensitive system information and environment variables, including 'AWS_SECRET_ACCESS_KEY', 'java.version', and 'hostName' to attacker-controlled external infrastructure.
- [REMOTE_CODE_EXECUTION]: Provides detailed payloads and methodological instructions for achieving remote code execution across various Java environments through RMI, LDAP, serialized gadget chains, and expression language (EL) injection.
- [COMMAND_EXECUTION]: Lists specific shell commands for initializing and running external exploitation frameworks such as 'marshalsec', 'ysoserial', and 'RogueJndi'.
- [PROMPT_INJECTION]: Uses 'AI LOAD INSTRUCTION' with authoritative framing ('Expert Attack Playbook', 'AI LOAD INSTRUCTION: Expert JNDI injection techniques') to direct the agent's behavior toward specialized offensive security tasks, which may circumvent standard safety guidelines regarding the generation of malicious payloads.
Audit Metadata