sd-navigate-pages
Pass
Audited by Gen Agent Trust Hub on Mar 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes browser automation tools to interact with ScienceDirect. It executes JavaScript via
evaluate_scriptto retrieve the current URL and page state. Additionally, it uses aninitScriptto mask thenavigator.webdriverproperty, which is a standard procedure to prevent bot detection. - [PROMPT_INJECTION]: The skill interprets user arguments to construct navigation URLs. This creates a surface for indirect prompt injection where untrusted input influences tool parameters.
- Ingestion points: User intent from
$ARGUMENTSis used to build target URL offsets and sort orders. - Boundary markers: No explicit delimiters are used to separate user input from the instruction context.
- Capability inventory: The skill possesses the capability to navigate to calculated URLs (
navigate_page) and execute scripts (evaluate_script). - Sanitization: The logic maps specific user intents (e.g., 'next', 'page 3') to predefined URL modifications, which reduces the risk of arbitrary injection.
Audit Metadata