zach-listing-health-checker
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The utility script
scripts/browser_utils.pyexecutes the systemcurlbinary viasubprocess.runto fetch Amazon listing and search web pages. The command is invoked safely by passing arguments as a list without shell expansion (shell=True), minimizing command injection risks. - [INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface because it retrieves and logs untrusted user-generated content from external Amazon product pages.
- Ingestion points: Public HTML data containing product titles and customer reviews is fetched and parsed within
scripts/fetch_amazon_page.pyandscripts/fetch_amazon_search.py. - Boundary markers: Absent; external data values are embedded directly into structured JSON and subsequent Markdown reports without specific isolation boundaries.
- Capability inventory: The tool possesses local file writing capabilities via Python file handlers and localized shell execution capabilities via
subprocess.runto invokecurl. - Sanitization: Text content is truncated to maximum length limits (e.g., reviews to 300 characters, titles to 120 characters), though no semantic filtering or escaping is implemented to sanitize text-based instructions.
Audit Metadata