skills/agentscope-ai/qwenpaw/news/Gen Agent Trust Hub

news

Pass

Audited by Gen Agent Trust Hub on Apr 28, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by processing content from external news websites.
  • Ingestion points: External news content is ingested into the agent context via the browser_use tool from several news domains.
  • Boundary markers: Absent. There are no explicit instructions to use delimiters or ignore instructions found within the fetched news content.
  • Capability inventory: The skill utilizes the browser_use tool to interact with web pages and extract content.
  • Sanitization: Absent. The agent extracts and summarizes content directly from the raw snapshots without validation.
  • [EXTERNAL_DOWNLOADS]: The skill directs the agent to fetch data from external URLs for news aggregation.
  • The skill references several established news outlets including People's Daily, China News, and CGTN.
  • One source, China Economic Net (http://www.ce.cn/), uses an unencrypted HTTP connection.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 28, 2026, 08:56 PM