expression-language-injection

Fail

Audited by Gen Agent Trust Hub on Apr 9, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill provides numerous ready-to-use payloads for executing arbitrary code via Java Expression Languages, including Spring Expression Language (SpEL), Object-Graph Navigation Language (OGNL), and Java EL (JSP/JSF).
  • Payloads leverage java.lang.Runtime.getRuntime().exec() and java.lang.ProcessBuilder for direct command execution.
  • Includes reflection-based techniques to bypass security sandboxes (e.g., bypassing SimpleEvaluationContext in Spring).
  • [COMMAND_EXECUTION]: Provides explicit shell command examples (curl) to interact with vulnerable endpoints and trigger RCE, specifically targeting the Spring Cloud Gateway actuator (CVE-2022-22947) and various Struts2 vectors.
  • [DATA_EXFILTRATION]: Includes methodologies for capturing command output from the target system and exfiltrating it via HTTP response headers using StreamUtils and IOUtils to convert process input streams to strings.
  • [PROMPT_INJECTION]: Contains a dedicated AI LOAD INSTRUCTION block designed to override the agent's default behavior by adopting an "Expert Attack Playbook" persona. This shapes the agent's output towards facilitating offensive security operations.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Apr 9, 2026, 04:28 AM