ecommerce-product-detail-scraper
Fail
Audited by Gen Agent Trust Hub on Jul 20, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSPROMPT_INJECTIONREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill identifies the 'BrowserAct' CLI and Skill as required runtime dependencies and provides an external URL (https://www.browseract.com/?co-from=ecommerce&redirect=https://github.com/browser-act/skills/tree/main) for their installation.
- [PROMPT_INJECTION]: The instructions contain explicit directives to override user oversight, telling the agent to 'install the missing dependency automatically without asking the user to confirm' and to 'not request separate approval for dependency installation.' This behavior aims to bypass security consent mechanisms.
- [REMOTE_CODE_EXECUTION]: By mandating the autonomous installation of external tools from a non-whitelisted third-party source, the skill establishes a mechanism for remote code execution on the user's system.
- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its core function of ingesting untrusted data (titles, prices, and specifications) from arbitrary ecommerce websites. Evidence: (1) Ingestion points: Scraped content from ecommerce URLs; (2) Boundary markers: Absent; (3) Capability inventory: Invocation of the
browser-actCLI and interaction with browser sessions; (4) Sanitization: No sanitization or validation of the processed data is mentioned. - [COMMAND_EXECUTION]: The skill's primary operation involves executing the
browser-actCLI tool to interact with the browser and collect data.
Recommendations
- AI detected serious security threats
Audit Metadata