news-rss-watch

Pass

Audited by Gen Agent Trust Hub on Aug 8, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill fetches content from a predefined registry of reputable news and government sources.
  • Sources include NPR, CNBC, WSJ, SEC, Federal Reserve, Google News, SCMP, BBC, The Guardian, and arXiv.
  • All network operations use standard HTTP GET requests via urllib.request with defined timeouts and byte-size limits to prevent resource exhaustion.
  • [COMMAND_EXECUTION]: Provides CLI scripts (watch_feeds.py and watch_arxiv.py) intended to be executed by the agent to perform news polling and search tasks.
  • The scripts are designed for deterministic output and use atomic file operations (os.replace) for state persistence.
  • [PROMPT_INJECTION]: The skill processes third-party news headlines and summaries, which constitutes an indirect prompt injection surface.
  • Ingestion points: Untrusted data enters the context via scripts/watch_feeds.py (RSS feeds) and scripts/watch_arxiv.py (arXiv Atom API).
  • Boundary markers: The output is structured into machine-readable JSON or Markdown with distinct headers, which helps delineate external content from instructions.
  • Capability inventory: The skill possesses the capability to write watermarks and digests to local directories (~/.finance-skills/) and emit content for agent consumption.
  • Sanitization: While titles are normalized for clustering, the skill does not perform specific filtering for instruction-like text within the news content. This is a low-risk inherent property of news-gathering tools.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 8, 2026, 11:46 AM
Security Audit — agent-trust-hub — news-rss-watch