alibabacloud-opensearch-app-manage
Pass
Audited by Gen Agent Trust Hub on Apr 24, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill provides instructions in 'SKILL.md' to download and execute an installation script from 'https://aliyuncli.alicdn.com/install.sh' via 'curl | bash'. This is the official distribution method for the author's own CLI tool.
- [EXTERNAL_DOWNLOADS]: Fetches CLI binary packages and installation scripts from official Alibaba Cloud domains ('alicdn.com').
- [COMMAND_EXECUTION]: Executes shell commands via the 'aliyun' CLI and standard system utilities like 'uuidgen' to manage search instance lifecycles.
- [PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it interpolates user-supplied data into shell commands. 1. Ingestion points: User-provided parameters such as 'instance name', 'spec', and 'type' in 'SKILL.md'. 2. Boundary markers: The skill explicitly instructs the agent to confirm all parameters with the user before executing any command. 3. Capability inventory: Execution of resource management commands via the 'aliyun' CLI. 4. Sanitization: Risk is mitigated by requiring manual user confirmation for all customizable values before the command is run.
Audit Metadata