israeli-heritage-explorer
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFECREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The file
evidence.jsoncontains a hardcoded guest API key (DVQyidFLOAjp12ib92pNJPmflmB5IessOq1CJQDK) intended for trial access to the National Library of Israel's API. While documented as a guest credential for public use, it remains a hardcoded identifier. - [INDIRECT_PROMPT_INJECTION]: The skill processes research data from the National Library of Israel's external digital archives. Maliciously crafted metadata within archival records could potentially influence agent behavior if not properly delimited.
- Ingestion points: The script
scripts/nli-search-example.pyfetches data from theapi.nli.org.ilendpoint. - Boundary markers: The instructions do not define explicit delimiters or instructions to ignore embedded commands within search results.
- Capability inventory: The skill performs network operations via
urllib.requestinscripts/nli-search-example.py. - Sanitization: The skill uses standard JSON parsing but lacks specific escaping or sanitization for the textual fields extracted from archival records before they are processed by the agent.
- [COMMAND_EXECUTION]: The skill includes a Python utility (
scripts/nli-search-example.py) and provides instructions for the agent to execute it to perform library searches.
Audit Metadata