skills/huaweicloud/huaweicloud-skills/huawei-cloud-modelarts-training-management/Gen Agent Trust Hub
huawei-cloud-modelarts-training-management
Pass
Audited by Gen Agent Trust Hub on Jul 28, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides instructions to download the hcloud CLI from the official Huawei Cloud support domain (
support.huaweicloud.com). These downloads are legitimate vendor resources required for the skill's functionality. - [REMOTE_CODE_EXECUTION]: Installation instructions include piping a
curlrequest from the vendor's official support site tobash. As this originates from the official author domain, it is considered an intended administrative operation. - [COMMAND_EXECUTION]: The skill translates user requests into
hcloudCLI commands. It includes a robust safety mechanism where all write operations (Create, Update, Delete, Stop) explicitly require the agent to obtain user confirmation before execution. - [DATA_EXFILTRATION]: Communication is directed to Huawei Cloud service endpoints for ModelArts API operations. These network operations are limited to the vendor's infrastructure and are necessary for the primary purpose of managing training jobs.
- [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection via user-supplied parameters (e.g., job names, descriptions) that are interpolated into CLI commands. This risk is effectively mitigated by the skill's instruction to the agent to confirm all commands with the user and the use of official, validated CLI tools.
Audit Metadata