ebay-search-listing
Pass
Audited by Gen Agent Trust Hub on Aug 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill employs dynamic script generation by using a Python script (scripts/extract-listing.py) to produce JavaScript code that is subsequently executed via eval. This pattern allows the agent to inject runtime-generated logic into the browser environment.
- [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it ingests untrusted data from external eBay pages.
- Ingestion points: The agent processes product titles, descriptions, and other text fields from eBay search and category result pages.
- Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the logic that handles scraped data.
- Capability inventory: The agent has permissions for browser navigation and JavaScript execution via the browser-act tool.
- Sanitization: While the extraction script cleans specific eBay UI strings and parses numeric values, it does not include filters or safety checks to prevent the agent from interpreting instructions contained within listing data.
Audit Metadata