vendor-verification

Pass

Audited by Gen Agent Trust Hub on Jun 28, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • External API Interaction: The skill performs GET requests to www.wikidata.org to search for entity information. This is an expected behavior for a verification tool and uses the official Wikidata API.
  • Local File System Access: The script reads from expenses.csv or reconciliation_data.json and writes results to a markdown report and a JSON file within the designated workspace. This file handling is standard for data processing tasks.
  • Command Execution: The skill involves executing a Python script via the shell to perform its logic, which is the primary mechanism for its functionality.
  • Indirect Prompt Injection Surface: The skill processes external vendor names found in CSV/JSON data. While it includes these names in its report, it does so as structured data or in tabular format, posing a low risk of influencing agent behavior.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 28, 2026, 06:45 AM
Security Audit — agent-trust-hub — vendor-verification