job-apply
Warn
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: MEDIUMDATA_EXFILTRATIONREMOTE_CODE_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The documentation provides instructions to install a local bridge service using remote scripts from Kimi's official CDN (cdn.kimi.com) via shell pipes.
- [DATA_EXFILTRATION]: Resume data containing personal information is written to a temporary JSON file ($env:TEMP\asu-webbridge-request.json) on the local disk to facilitate bridge communication.
- [DYNAMIC_EXECUTION]: The skill utilizes browser-level actions such as 'evaluate' and 'cdp' (Chrome DevTools Protocol) to interact with complex web elements, allowing for arbitrary code execution within the browser session.
- [INDIRECT_PROMPT_INJECTION]: The skill processes DOM and accessibility snapshots from external recruitment websites to determine automation steps. Evidence chain: 1. Ingestion: 'snapshot' action from recruitment pages. 2. Boundaries: Instructions to treat page content as untrusted. 3. Capabilities: Actions include 'upload', 'click', and 'evaluate'. 4. Sanitization: No specific data sanitization mentioned.
- [EXTERNAL_DOWNLOADS]: Fetches configuration and setup components from the Kimi official domain.
- [COMMAND_EXECUTION]: Executes local Node.js scripts to manage the connection status and transmit commands to the browser bridge.
Audit Metadata