dashscope
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads generated images and audio files from Alibaba Cloud's DashScope API (dashscope.aliyuncs.com) and Zhipu AI's API (open.bigmodel.cn). These are official endpoints for well-known technology services.
- [COMMAND_EXECUTION]: Documentation includes examples of using
curlto interact with REST APIs andbase64to process audio files for voice cloning tasks. - [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it processes external inputs (text prompts and image URLs) that are then passed to external AI models.
- Ingestion points: Processes user-provided
prompt,text, andimage_urlfields across SKILL.md and reference files. - Boundary markers: No explicit delimiters or boundary markers are used in the prompt construction logic shown in the documented snippets.
- Capability inventory: Writing downloaded media files to the local file system using
open()calls in Python code examples. - Sanitization: Relies on the external API provider's internal safety filters.
- [DATA_EXPOSURE]: The skill handles local file paths such as
/tmp/and/home/ubuntu/hermes/cache/for storing temporary media assets. It correctly recommends using.envfiles for managing API keys, which is a standard security best practice.
Audit Metadata