huawei-cloud-deployment-task-management
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSDATA_EXFILTRATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill's setup documentation provides instructions for downloading the hcloud CLI from a Huawei Cloud OBS bucket (cn-north-4-hcli.obs.cn-north-4.myhuaweicloud.com). This is a well-known service endpoint belonging to the vendor and is used only for initial environment preparation.
- [DATA_EXFILTRATION]: The skill incorporates a telemetry script (skill_quality_sdk.py) that reports execution metadata, including trace IDs and masked inputs/outputs, to a Huawei Cloud quality monitoring endpoint (skillsapi.developer.myhuaweicloud.com). The script implements robust sanitization logic using regex patterns to mask access keys, secret keys, passwords, and tokens before transmission. Since both the skill and the endpoint are managed by the vendor (huaweicloud), this represents standard operational quality assurance.
- [COMMAND_EXECUTION]: The skill executes cloud management tasks through the hcloud CLI. It strictly adheres to confirmation protocols (R1 for delete, R2 for create/start), ensuring that no destructive or resource-intensive commands are run without a human-in-the-loop preview and approval.
- [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from cloud logs and task configurations.
- Ingestion points: Task details and history logs retrieved via hcloud ShowDeployTaskDetail and ListDeployTaskHistoryByDate.
- Boundary markers: Output is reported as structured JSON; state-changing commands are isolated by manual confirmation steps.
- Capability inventory: Cloud resource management actions (CreateApp, StartDeployTask, DeleteDeployTask) via hcloud CLI.
- Sanitization: Sensitive data is masked by the quality SDK; write operations require explicit human confirmation.
Audit Metadata