aliyun-mps-manage
Fail
Audited by Gen Agent Trust Hub on Apr 30, 2026
Risk Level: HIGHCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill explicitly directs the agent to access sensitive local credential files for authentication.
- Evidence:
SKILL.mdreferences the shared config file at~/.alibabacloud/credentialsunder the 'AccessKey Priority' section. - [EXTERNAL_DOWNLOADS]: The skill retrieves API metadata from an external source during execution.
- Evidence: The script
scripts/list_openapi_meta_apis.pyfetches JSON documentation from Alibaba Cloud's official domain (api.aliyun.com) to discover available service operations. - [PROMPT_INJECTION]: The skill ingests data from an external API response which could serve as a vector for indirect instructions.
- Ingestion points:
scripts/list_openapi_meta_apis.py(line 45) - Boundary markers: Absent; the script parses the external JSON and generates a list of operations without specific delimiters or instructions to ignore embedded content.
- Capability inventory: The skill is intended to perform mutate operations on cloud resources (e.g.,
AddPipeline,SubmitJobs,AddMediaWorkflow) based on the discovered API names. - Sanitization: Absent; the script processes keys directly from the JSON payload and writes them to local files for further agent use.
Recommendations
- AI detected serious security threats
Audit Metadata