rednote-skill

Pass

Audited by Gen Agent Trust Hub on Feb 25, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill operates by executing a series of standalone Python scripts (e.g., publish_note.py, comment_note.py) that utilize the Playwright library to automate browser tasks.\n- [DATA_EXFILTRATION]: The publish_note.py script includes functionality to upload local files to the Xiaohongshu platform via the --image-urls parameter. This is a core feature for publishing content, though it creates an interface for accessing local files.\n- [PROMPT_INJECTION]: The skill consumes content from external note URLs and keywords. This content is untrusted and may contain instructions targeting the agent. \n
  • Ingestion points: dump_note.py and search_note_by_key_word.py. \n
  • Boundary markers: No delimiters or instructions to ignore embedded commands were identified in the scripts. \n
  • Capability inventory: The skill has capabilities to interact with the platform (commenting, following, publishing) based on processed data. \n
  • Sanitization: No sanitization or validation of the extracted content is performed.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 25, 2026, 08:46 AM