story-long-scan
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONOBFUSCATIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill relies on the
agent-browserCLI tool to control a browser for web scraping. This is implemented inscripts/cdp-utils.jsusingexecFileSyncto interact with the external binary. This execution is a core part of the skill's data collection functionality. - [OBFUSCATION]: In
scripts/cdp-utils.js, theevalJSONBase64function encodes JavaScript scraping payloads into Base64 before passing them to the browser controller. As documented in the code, this is a technical measure to prevent character escaping issues across different operating system shells (such as Windows cmd.exe) rather than a malicious attempt to hide intent. - [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes large amounts of untrusted content from external novel platforms, including book titles, summaries, and tags. While the
cleanDescfunction performs basic text normalization and truncation, the ingestion of external data for market analysis inherently presents a surface for potential indirect prompt injection attacks. - [DYNAMIC_EXECUTION]: The scrapers dynamically generate JavaScript snippets at runtime to extract structured data from web pages. These scripts are then evaluated within the browser context via the Chrome DevTools Protocol (CDP) interface provided by the underlying browser controller.
Audit Metadata