NYC

article-extractor

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [Unverifiable Dependencies] (LOW): The skill installs @mozilla/readability-cli (trusted source) and trafilatura (external repository) packages to perform its core function. These dependencies represent external code execution risks.\n- [Indirect Prompt Injection] (LOW): The skill ingests untrusted content from arbitrary URLs, creating an attack surface where malicious web pages could attempt to influence agent behavior through hidden instructions.\n
  • Ingestion points: External web content via the ARTICLE_URL variable.\n
  • Boundary markers: Absent; the skill saves and previews extracted text without using delimiters or system-level isolation.\n
  • Capability inventory: The skill uses Bash (for system commands) and Write (for file system access).\n
  • Sanitization: Web content is parsed for text but not sanitized for embedded instructions that could target the LLM.\n- [Dynamic Execution] (LOW): Uses python3 -c for parsing and interpolates variables directly into shell commands. While localized and necessary for the primary purpose, this pattern carries a minor risk of command injection if the input URLs or titles are specifically crafted to exploit shell syntax.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 05:59 PM