alibabacloud-dts-job-manager
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches the
dtsclimanagement tool installer from Alibaba Cloud's official storage service (aliyun-dts-cli.oss-cn-hangzhou.aliyuncs.com). This is a legitimate download from the vendor's infrastructure for tool setup. - [REMOTE_CODE_EXECUTION]: The skill executes the official installation scripts (
install.shon macOS/Linux andinstall.ps1on Windows) to configure the local runtime environment. As these scripts originate from a trusted vendor domain, they are considered safe for their intended purpose. - [COMMAND_EXECUTION]: The skill makes extensive use of the
dtsclicommand-line utility to perform cloud operations, including instance resolution, price querying, and job lifecycle management. All management actions (start, pause, rename, execute) require explicit, recent user confirmation. - [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect injection as it processes data from external cloud APIs (e.g., job lists and status details). It mitigates this risk with clear instructions to redact sensitive information, validate all inputs against an internal schema, and require human review for plan changes.
- Ingestion points: Job metadata and status queried via
dtscli job listanddtscli job get(documented injob-status.md). - Boundary markers: The skill uses clear delimiters for plan review and requires explicit user approval before execution.
- Capability inventory: The skill can write files (plans), perform network requests via the CLI, and manage cloud job states.
- Sanitization: The skill implements redaction for usernames, failure diagnostics, and internal identifiers before displaying them to the user.
- [CREDENTIALS_UNSAFE]: The skill describes the local storage of database credentials in Base64 format under the
.aliyun-dts/job-manager/credentials/directory. It explicitly instructs the agent to never accept secrets in chat and uses a browser-based collection flow provided by the CLI to ensure security.
Audit Metadata