mini-six-ren
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill relies on standard, well-known Python libraries including
lunardate,pydantic-ai-slim, andpyyamlfor its core functionality. These dependencies are typical for AI-integrated tools. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied text (questions and Chinese characters) and includes them in prompts sent to LLMs for analysis. While this presents a surface for indirect prompt injection, the scope is limited to the divination response and does not provide access to sensitive system resources or tools.
- [COMMAND_EXECUTION]: The workflow involves running local Python scripts (
xiaoliu.pyandinterpret.py) using theuvpackage manager. These scripts implement the divination logic and LLM orchestration as described in the documentation and do not perform unauthorized system modifications or privilege escalation. - [DATA_EXFILTRATION]: Network activity is confined to official API endpoints of established LLM providers (e.g., OpenAI, Anthropic, DeepSeek, Google) for the sole purpose of interpreting divination results. No evidence of unauthorized data harvesting or transmission to untrusted third parties was found.
Audit Metadata