story-short-scan
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes the
agent-browserCLI tool usingchild_process.execFileSyncwithinscripts/cdp-utils.jsto facilitate browser-based automation tasks.\n- [DYNAMIC_EXECUTION]: Scraper scripts inscripts/dz-browse-scraper.jsandscripts/heiyan-booklist-scraper.jsgenerate JavaScript code at runtime and execute it in the browser context usingeval(viaagent-browser) to perform data extraction and API calls.\n- [INDIRECT_PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection by ingesting and processing untrusted content from external web platforms.\n - Ingestion points: Data is extracted from
ishugui.comand collected via API responses fromms.zhangwenpindu.cn.\n - Boundary markers: No explicit delimitation or sanitization warnings are present for the ingested data.\n
- Capability inventory: The skill possesses file-write capabilities via
fs.writeFileSyncand command execution viaexecFileSync.\n - Sanitization: Scraped text fields, including book titles and descriptions, are interpolated directly into Markdown reports without escaping or filtering.\n- [DATA_EXFILTRATION]: The script
scripts/heiyan-booklist-scraper.jsretrieves theAdmin-Tokenfrom the browser's cookies. This sensitive token is subsequently used in authenticatedfetchrequests to the platform's backend athttps://ms.zhangwenpindu.cnto retrieve book lists and details, which is the primary intended function of the skill for the Heiyan platform.
Audit Metadata