crawl-xueqiu-my-timeline
Pass
Audited by Gen Agent Trust Hub on Mar 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes multiple shell scripts and Python programs to manage the local environment. It utilizes
pkillto terminate existing 'Google Chrome' processes and launches Chromium with specific debugging flags (--remote-debugging-port=9222). - [EXTERNAL_DOWNLOADS]: Installation scripts dynamically download and install external dependencies, including
node@22via Homebrew and theagent-browserpackage from the well-knownnpmmirror.comregistry. - [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it processes untrusted user-generated content from Xueqiu social media feeds for AI analysis.
- Ingestion points: External text data enters the context via the
crawl_xueqiu_home_timeline_api.pyscript, which fetches posts from the Xueqiu API. - Boundary markers: There are no explicit markers or instructions to isolate the crawled text from the AI's core instructions, making it possible for posts to contain malicious prompts.
- Capability inventory: The skill possesses capabilities for shell command execution, filesystem writes, and browser manipulation via debugging protocols.
- Sanitization: The skill performs basic HTML tag stripping but lacks semantic sanitization to filter out potential instructions embedded in the social media posts.
Audit Metadata