database-tactics

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt includes explicit plaintext passwords and credential templates embedded in commands (e.g., netexec/impacket with sa/PASSWORD, common default passwords, and brute-force examples), which requires the LLM to output secret values verbatim and thus is insecure credential handling.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The content is an explicit, actionable offensive guide for unauthorized database access, credential theft, RCE, and persistence (writing webshells/SSH keys/crontab), indicating intentional malicious intent and abuse potential.

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

• Third-party content exposure detected (medium risk: 0.30). 该技能运行路径是在回答评测/用户提问时摄取“prompt”中的免费文本（例如 evals/evals.json 的 prompt）并据此生成数据库入侵步骤，因此存在被外部文本诱导的间接提示注入暴露。

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The skill explicitly instructs unauthorized access, weak-password brute force, command execution and file writes (e.g., "写 Webshell / SSH 密钥 / Crontab") and guides OS privilege escalation on target systems, which directly encourages modifying machine state and system compromise.