alibabacloud-ram-permission-diagnose

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill tells the agent to transcribe the EncodedDiagnosticMessage from an error and insert it verbatim into a CLI call (aliyun ram decode-diagnostic-message), which forces the LLM to handle and output a potentially sensitive value directly.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs running a remote install script that is fetched and executed via "curl -fsSL https://aliyuncli.alicdn.com/setup.sh | bash", which downloads and runs remote code as part of setup.