shiliu
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches supplementary documentation, API schemas, and third-party creative skill definitions from the official vendor domain
shiliuai.cn. These resources are used to guide the agent's research and writing tasks. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, including user-supplied account links, platform search results, and third-party skill definitions. This presents an indirect prompt injection surface. However, the skill explicitly instructs the agent to treat these external sources strictly as data or 'source material' and forbids them from overriding user instructions or granting unauthorized access.
- [CREDENTIALS_UNSAFE]: The skill identifies the need for API keys for private operations but directs the agent to use secure client-side secret configuration. It explicitly prohibits the user from providing secrets via public conversation, URL parameters, or other insecure methods.
Audit Metadata