The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides commands to download installation scripts for 'uv' (from astral.sh) and the Alibaba Cloud CLI (from aliyuncli.alicdn.com). These downloads originate from well-known or vendor-managed sources and are required to configure the skill's operating environment.
[REMOTE_CODE_EXECUTION]: Contains patterns for fetching and executing shell scripts via pipes or subshells. These patterns are limited to initial setup instructions targeting trusted vendors and are not executed autonomously with arbitrary inputs.
[COMMAND_EXECUTION]: Utilizes local subshell execution to interact with the Alibaba Cloud CLI and manage the MCP proxy server's lifecycle. This is the primary intended behavior of the connector.
[DYNAMIC_EXECUTION]: Provides the 'RunScript' and 'RunIaC' tools, which allow for the execution of Python and Terraform code within a managed server-side sandbox. These capabilities are designed for legitimate infrastructure orchestration and batch operations.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from cloud service API responses and official documentation.
Ingestion points: Data from Alibaba Cloud APIs and official product documentation is ingested into the agent context during cloud operations.
Boundary markers: Employs structured JSON outputs and JMESPath filtering to maintain separation between retrieved data and agent instructions.
Capability inventory: Extensive capabilities for cloud resource management (e.g., ECS, OSS, VPC), script execution, and Terraform orchestration.
Sanitization: Prioritizes official Alibaba Cloud document search tools over general web scraping, ensuring that retrieved information is authoritative and safe.

alibabacloud-mcp-connector