alibabacloud-agentloop-dataset
Pass
Audited by Gen Agent Trust Hub on Jul 17, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the aliyun CLI to manage cloud resources, which is its intended purpose. It includes safeguards like using --cli-dry-run to inspect commands before execution.
- [CREDENTIALS_UNSAFE]: The skill adheres to best practices by explicitly forbidding the handling of literal secrets. It provides clear instructions to use authenticated CLI profiles and avoid echoing or printing credentials.
- [PROMPT_INJECTION]: The skill uses benign instructional language and does not contain any patterns intended to override agent safety or reveal system prompts.
- [REMOTE_CODE_EXECUTION]: No remote code execution or suspicious download patterns were detected. It correctly directs users to official package management commands (aliyun plugin install) for necessary extensions.
- [SAFE]: The skill identifies the surface where indirect prompt injection could occur and provides mitigation strategies: 1) Ingestion points: Data rows appended to datasets and query parameters. 2) Boundary markers: Instructions emphasize building queries from documented patterns rather than untrusted strings and recommend dry runs. 3) Capability inventory: The skill uses aliyun agentloop to create, write, and query datasets. 4) Sanitization: Data is validated against defined schemas before being committed.
Audit Metadata