xiaohongshu-account-recommender

Warn

Audited by Gen Agent Trust Hub on May 15, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The script scripts/xiaohongshu_account_recommender.py fetches account data from https://onetotenvip.com/story/xhsUser/querySimilarAccounts using an unverified HTTPS connection.
  • [COMMAND_EXECUTION]: The skill executes a Python script that performs raw socket network operations and writes data to files such as account_data.json and account_recommend.html within the workspace.
  • [PROMPT_INJECTION]: The instructions in SKILL.md require the agent to read and display a local HTML file generated from external API data, which allows for potentially malicious content (like scripts or HTML injection) to be processed and rendered by the agent platform.
  • [INDIRECT_PROMPT_INJECTION]:
  • Ingestion points: Data is ingested from an external API endpoint via scripts/xiaohongshu_account_recommender.py.
  • Boundary markers: None; the skill does not use delimiters or instructions to ignore potential commands within the external data.
  • Capability inventory: The skill has file-write access and network access; the agent is instructed to display the generated local HTML file.
  • Sanitization: Absent; the Python script directly interpolates raw data from the API response (such as nicknames and recommendation reasons) into the HTML template without escaping special characters.
Audit Metadata
Risk Level
MEDIUM
Analyzed
May 15, 2026, 11:01 PM
Security Audit — agent-trust-hub — xiaohongshu-account-recommender