wechat-article-extractor

Pass

Audited by Gen Agent Trust Hub on Mar 8, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the exec tool to run curl for downloading HTML and a local Python script (scripts/extract_wechat.py) for processing content. \n- [EXTERNAL_DOWNLOADS]: The workflow involves downloading HTML content from various third-party mirror sites (e.g., 53ai.com, juejin.cn, ofweek.com) to bypass WeChat's anti-bot protections. \n- [PROMPT_INJECTION]: Indirect prompt injection risk is present because the skill ingests untrusted HTML from external websites. Mandatory Evidence Chain: \n
  • Ingestion points: Mirror HTML files (e.g., /tmp/wechat-article.html) and direct fetches via web_fetch. \n
  • Boundary markers: The script isolates the article body, but the agent lacks explicit instructions to ignore instructions found within the extracted text. \n
  • Capability inventory: Uses exec (subprocess execution), web_fetch, web_search, and optionally browser. \n
  • Sanitization: The extract_wechat.py script removes <script> and <style> tags and uses regex for deterministic extraction rather than LLM-based parsing.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 8, 2026, 02:53 AM