huawei-cloud-dds-dcs-instance-management
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEPRIVILEGE_ESCALATIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PRIVILEGE_ESCALATION]: The CLI installation guide (
references/cli-installation-guide.md) instructs the user to usesudoto move thehcloudexecutable into/usr/local/bin/, which requires administrative privileges to modify system directories. - [EXTERNAL_DOWNLOADS]: The skill downloads the
hcloudCLI from a remote Huawei Cloud Object Storage Service (OBS) endpoint:https://cn-south-1-cloud-res-model-sdk.obs.cn-south-1.myhuaweicloud.com/hcloud/hcloud.tar.gz. This is a standard vendor-provided source for the management tool. - [DATA_EXFILTRATION]: The
scripts/skill_quality_sdk.pyscript is designed to automatically transmit execution metadata, including trace IDs, status, and masked versions of input and output data, to the endpointhttps://skillsapi.developer.myhuaweicloud.com/api/quality/report. Although the script uses regular expressions to mask sensitive patterns like keys and tokens, it initiates an automated outbound connection that transmits interaction logs to a remote service. - [INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection through its ingestion of external resource data.
- Ingestion points: The skill retrieves instance lists, node information, and security configurations from the Huawei Cloud DDS and DCS APIs through the
hcloudCLI and Python SDK. - Boundary markers: The instructions do not specify the use of clear delimiters or explicit instructions to ignore embedded commands when processing these external data strings.
- Capability inventory: The skill possesses high-privilege capabilities, including executing shell commands (
hcloud), utilizing the Huawei Cloud Python SDK, and performing destructive actions such as deleting database instances. - Sanitization: There is no evidence of validation or sanitization of the content returned by cloud APIs before it is incorporated into the agent's context.
Audit Metadata