douyin-search
Warn
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPERSISTENCEINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses
python3to execute a local script (search_douyin.py) with user-supplied search terms as arguments, facilitating core search functionality via the shell.\n- [PERSISTENCE]: A daily subscription feature is implemented by adding entries to the systemcrontab, allowing for automated daily execution of search commands. This persistence mechanism requires careful handling of user input to prevent unauthorized command scheduling.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests and displays external video titles and author data, which could be exploited to deliver malicious instructions to the agent.\n - Ingestion points: Data is fetched from the search API in
scripts/search_douyin.pyand returned as thearticleslist containing user-controlled content from Douyin.\n - Boundary markers: The instructions lack explicit requirements for delimiters or specific instructions to ignore embedded content when rendering video titles in the results table.\n
- Capability inventory: The skill has the capability to perform shell command execution and modify system persistence files (crontab).\n
- Sanitization: No sanitization or filtering of external video metadata is specified before it is presented to the user or processed by the agent.
Audit Metadata