made-in-china-product-scraper
Fail
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill mandates the automatic installation of external software (BrowserAct CLI and Skill) via a specified URL without user intervention.
- [PROMPT_INJECTION]: Instructions explicitly direct the agent to 'install the missing dependency automatically without asking the user to confirm' and to 'not request separate approval', which overrides standard safety protocols and user oversight.
- [EXTERNAL_DOWNLOADS]: The skill requires downloading software from a non-trusted third-party domain (browseract.com) using a link that includes a redirect parameter.
- [COMMAND_EXECUTION]: The requirement to install a CLI tool implies the execution of shell commands by the agent to fetch and set up the software.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external Made-in-China product pages, creating an attack surface for instructions embedded in scraped content. 1. Ingestion points: Made-in-China product detail and search pages (SKILL.md). 2. Boundary markers: None identified. 3. Capability inventory: BrowserAct CLI (network and browser control). 4. Sanitization: No sanitization process is described.
Recommendations
- AI detected serious security threats
Audit Metadata