The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill clones and analyzes external Git repositories or local project directories to automate the deployment process. This creates a vulnerability surface where malicious files in the source repository could potentially influence the agent's behavior.
Ingestion points: The skill reads project metadata, configuration files (e.g., "package.json", "requirements.txt"), and entry source files during the analysis phase.
Boundary markers: No explicit delimiters or instructions to ignore embedded instructions in the analyzed data were identified.
Capability inventory: The skill uses the "aliyun" CLI for cloud resource orchestration and the Cloud Assistant for remote command execution on cloud instances.
Sanitization: The tool includes a check for hardcoded secrets but does not implement specific measures to sanitize or escape project content to prevent prompt injection.
[COMMAND_EXECUTION]: The skill makes extensive use of the "aliyun" CLI and "docker" commands to manage cloud infrastructure and application containers. This includes operations like stack creation, OSS bucket management, and DNS configuration.
[REMOTE_CODE_EXECUTION]: The skill leverages the Alibaba Cloud Assistant ("aliyun ecs RunCommand") to execute scripts on the deployed ECS instances. This is a core feature used for runtime installation, Nginx/HTTPS setup, and atomic application updates.
[EXTERNAL_DOWNLOADS]: The skill downloads binaries and dependencies from external sources. It fetches the Docker Compose binary from the official Docker GitHub repository and uses system package managers to install runtimes like Node.js, Java, and Python.

qwencloud-deploy