douyin-search
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes a Bash script (
scripts/douyin_task.sh) to executecurlandjqfor API interactions. This is the primary intended behavior and includes thorough validation of user-provided arguments. - [EXTERNAL_DOWNLOADS]: The script connects to the Frevana API at
ai-factory.frevana.comto search for and retrieve Douyin video metadata. This is the service provider required for the skill's functionality. - [INDIRECT_PROMPT_INJECTION]: As the skill fetches and returns data from an external social media platform (Douyin), it constitutes a surface for indirect prompt injection if the agent processes the resulting text content.
- Ingestion points: Data enters via the
resultaction inscripts/douyin_task.shwhich fetches JSON results from the Frevana API. - Boundary markers: The skill documentation does not define specific delimiters for the search results in the agent's context.
- Capability inventory: The script allows network requests via
curland file writing via the--outputflag. - Sanitization: The script uses
jqto ensure valid JSON structure for API payloads and results, but does not sanitize the content of the retrieved search items.
Audit Metadata