ecommerce-google-ai-mode-search
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process unstructured Markdown content from external Google Search AI Overviews, which creates a surface for indirect prompt injection.\n
- Ingestion points: External content is received via the
stdoutfield of the API response inscripts/google_ai_search.pyand processed inSKILL.md.\n - Boundary markers: The instructions lack explicit boundary markers or 'ignore' directives when displaying the retrieved Markdown to the user or agent.\n
- Capability inventory: The skill executes shell commands via
scripts/google_ai_search.py, performs network requests to an external API, and writes JSON files to the local filesystem (os.getcwd()/nexscope).\n - Sanitization: No sanitization or validation of the retrieved web content is performed before it is used to construct subsequent search queries or rendered in the agent's context.\n- [COMMAND_EXECUTION]: The skill relies on a Python script (
scripts/google_ai_search.py) that is executed via the command line with JSON-encoded parameters. This involves the execution of shell commands where user-provided keywords are passed as arguments.
Audit Metadata