smart-search
Pass
Audited by Gen Agent Trust Hub on Nov 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the
camoufox-clitool to perform browser automation, such as navigating to search URLs, capturing snapshots, and managing persistent user sessions. - [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from a wide variety of external web sources, which is a known vector for indirect attacks.
- Ingestion points: Web page titles, summaries, and full-text content are extracted from the search results of numerous platforms, including Bing, Bilibili, YouTube, and GitHub.
- Boundary markers: The instructions do not mandate the use of clear delimiters or specific protective instructions when the agent incorporates retrieved web content into its context.
- Capability inventory: The skill allows the agent to execute shell commands via the
camoufox-clitool and interact with local browser profiles. - Sanitization: There are no requirements for sanitizing or filtering the scraped web content before it is processed by the agent's language model.
- [DYNAMIC_EXECUTION]: The instructions direct the agent to use
evalwithin the browser's context to extract data from the Document Object Model (DOM) when standard extraction methods are insufficient.
Audit Metadata