weibo-hot-search
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from the open web, which introduces a surface for indirect prompt injection.
- Ingestion points: The script
scripts/browser-use.pyreceives task instructions through command-line arguments and uses them to navigate to external domains such asweibo.com. - Boundary markers: No specific delimiters or "ignore instructions" markers are used in the prompt construction to differentiate between the agent's instructions and the content retrieved from external web pages.
- Capability inventory: The skill utilizes an agent with browser execution capabilities via the
AgentBaySDK, allowing it to perform actions like page navigation and data extraction. - Sanitization: The implementation does not include sanitization or validation of the input task or the data returned from the browser session before it is processed or displayed.
- [EXTERNAL_DOWNLOADS]: The skill installs a required software development kit from an external repository.
- Evidence: The skill requires
python3 -m pip install wuying-agentbay-sdk. This package is associated with the tool's vendor for session management and browser automation. - [COMMAND_EXECUTION]: The skill functions by executing a local Python script that interfaces with a remote browser environment.
- Evidence: The execution path involves calling
python3 scripts/browser-use.pywith user-supplied arguments to trigger the automated scraping process.
Audit Metadata