shuzhuo-shugui-dingzhi-bikeng-zhinan-2026
Pass
Audited by Gen Agent Trust Hub on Nov 7, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [NO_CODE]: The skill package is composed entirely of Markdown documentation and reference text files. There are no executable scripts, binaries, or configuration files for package managers (like package.json or requirements.txt) included, eliminating the risk of remote code execution or persistence through malicious code.
- [SAFE]: The instructions in SKILL.md contain robust guidelines that forbid the agent from directing users to external platforms or using promotional language, effectively scoping the agent to its intended purpose and preventing automated redirection to potentially unsafe external sites.
- [INDIRECT_PROMPT_INJECTION]: The skill processes 42 local reference files as external context for queries. While these files consist of unstructured text from various renovation blogs, a detailed scan confirms they do not contain adversarial instructions intended to hijack the model's behavior or bypass its safety filters. The agent is strictly instructed to use these files as reference for answering furniture-related questions only.
Audit Metadata