alibabacloud-data-agent-skill
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses subprocesses to manage background execution and local environment integrations. In
scripts/cli/worker_utils.py,subprocess.Popenis used to spawn a worker process by re-executing the current Python script with modified environment variables (DATA_AGENT_ASYNC_WORKER=1). Inscripts/cli/notify.py,subprocess.runis used to detect and interact with local notification tools likeopenclaworclawdbotto provide status updates. Inscripts/cli/dual_logger.py,subprocess.Popenis used to capture command output for logging purposes. - [EXTERNAL_DOWNLOADS]: The skill downloads agent-generated reports and charts from Alibaba Cloud's Object Storage Service (OSS). In
scripts/cli/cmd_db.pyandscripts/cli/cmd_reports.py, the skill fetches download URLs for generated files (e.g., WebReports) and saves them locally using theFileManager. These downloads target official Alibaba Cloud storage infrastructure associated with the Data Agent service. - [INDIRECT_PROMPT_INJECTION]: The skill serves as a surface for indirect prompt injection as it processes data from external databases and files. Ingestion points: Data is ingested via the
dbandfilecommands which read metadata and content from connected Alibaba Cloud databases and local files (scripts/cli/cmd_db.py,scripts/cli/cmd_file.py). Boundary markers: The client-side skill does not explicitly wrap ingested data in delimiters or append instructions to ignore embedded commands before sending the data to the server-side agent. Capability inventory: The skill has capabilities to perform network API calls to Alibaba Cloud services, write session artifacts to the local filesystem, and spawn background worker processes. Sanitization: The skill does not perform client-side sanitization of the database content or file data being analyzed. - [CREDENTIALS_SAFE]: The skill demonstrates safe credential handling practices. It utilizes the official Alibaba Cloud default credential chain, which supports secure methods like environment variables, local configuration files (
~/.aliyun/config.json), or instance roles. Inscripts/data_agent/client.py, the skill implements a redaction mechanism (_redact_sensitive_values) that masks fields matching sensitive patterns (e.g., 'api-key', 'secret', 'token') before they are written to debug logs.
Audit Metadata