research-assistant
Pass
Audited by Gen Agent Trust Hub on Apr 18, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill includes multiple prompts intended to bypass AI-generated content (AIGC) detection filters, such as those found in 'references/03-paper-writing.md' (items 20 and 28) and the '降低AIGC Prompt' section in 'references/05-paper-prompts.md'. These instructions guide the agent to manipulate writing styles specifically to evade automated detection tools.
- [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by processing untrusted external manuscripts and PDFs (e.g., 'references/03-paper-writing.md' items 15, 18, and 45). [Ingestion points]: User-provided manuscripts, PDF documents, and web search results. [Boundary markers]: Absent; the prompts do not utilize delimiters or instructions to isolate external text from system logic. [Capability inventory]: High capability to summarize, rewrite, and interpret content which can be manipulated by malicious input within processed documents. [Sanitization]: No validation or sanitization of external content is performed.
Audit Metadata