skills/zhuxining/harness/news/Gen Agent Trust Hub

news

Pass

Audited by Gen Agent Trust Hub on Apr 25, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by ingesting and processing content from external, untrusted web sources.
  • Ingestion points: Data enters the agent's context through the search-and-fetch tool, as well as via web crawlers like defuddle and crwl which fetch content from URLs defined in config/sources.yaml and dynamic search results.
  • Boundary markers: The skill lacks explicit instructions or delimiters (e.g., XML tags or clear 'ignore instructions' warnings) to encapsulate fetched news content and prevent it from being interpreted as a command by the LLM.
  • Capability inventory: The skill utilizes web search and navigation capabilities (search-and-fetch, defuddle, crwl, agent-browser) to interact with external environments.
  • Sanitization: There is no logic present to sanitize or filter the content of news summaries for potentially malicious instructions or prompt injection patterns before they are displayed to the user.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 25, 2026, 05:01 AM