fetching-ai-news

Pass

Audited by Gen Agent Trust Hub on May 20, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill fetches data from well-known services including the Hacker News Algolia API, ArXiv RSS feeds, and Google News RSS feeds to aggregate news items.\n- [DATA_EXFILTRATION]: Provides the ability to send formatted news digests to the Telegram Bot API. This behavior is documented and requires user-provided credentials via environment variables.\n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted content retrieved from external sources.\n
  • Ingestion points: External data is fetched in references/src/sources/hackernews.py, references/src/sources/arxiv.py, and references/src/sources/googlenews.py.\n
  • Boundary markers: The output is structured using numbering and markdown in references/src/formatter.py, but lacks explicit 'ignore instructions' delimiters for the external content.\n
  • Capability inventory: The skill has the capability to send data to the Telegram API (references/src/telegram.py) and returns raw fetched content to the agent via an MCP tool (references/src/mcp_server.py).\n
  • Sanitization: The references/src/processor.py module performs basic HTML tag removal on summaries, but does not sanitize the semantic content for potential malicious instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
May 20, 2026, 03:55 PM
Security Audit — agent-trust-hub — fetching-ai-news