alibabacloud-oss-manage-metaquery
Pass
Audited by Gen Agent Trust Hub on May 2, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides instructions to download and install official Alibaba Cloud CLI and ossutil binaries from vendor-controlled domains including
aliyuncli.alicdn.comandgosspublic.alicdn.com. These are recognized as legitimate well-known service endpoints. - [COMMAND_EXECUTION]: The skill uses the
aliyunCLI and custom Python scripts to perform storage management tasks. It correctly identifies dangerous operations—such as bucket deletion (aliyun ossutil rm -b) or metadata index closure (aliyun ossutil api close-meta-query)—and mandates explicit user confirmation before execution to prevent accidental data loss. - [PROMPT_INJECTION]: The skill is identified as having an architectural surface for indirect prompt injection (Category 8).
- Ingestion points: The agent ingests AI-generated metadata (descriptions and summaries) from OSS objects when executing the
do-meta-querycommand, as detailed inSKILL.mdandreferences/verification-method.md. - Boundary markers: There are no specific delimiters or instruction-ignore warnings applied to the AI-generated metadata before it is displayed to the user.
- Capability inventory: The agent possesses capabilities to create/delete buckets and manipulate objects via the
aliyuntoolset and provided Python scripts (e.g.,scripts/create_bucket.py). - Sanitization: No explicit sanitization or filtering logic is implemented for the text content returned in the
oss_ai_metafields before it enters the agent's conversational context.
Audit Metadata