byted-viking-aisearch-database

Pass

Audited by Gen Agent Trust Hub on Jul 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill enables the agent to execute SQL commands, including data modification and schema changes via the execute_sql tool. The provided instructions emphasize safe operation and require user confirmation for any data-altering actions.
  • [EXTERNAL_DOWNLOADS]: The skill requires the 'volcengine' and 'pandas' Python packages. The 'volcengine' package is the official SDK for the vendor's cloud services.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through database query results.
  • Ingestion points: Data retrieved from database tables and metadata via the execute_sql and get_table_info methods in scripts/tunnel.py.
  • Boundary markers: The instructions guide the agent to verify results, but the implementation does not use technical delimiters or markers to isolate ingested external data.
  • Capability inventory: The skill can perform broad database operations, which could be abused if malicious instructions in database records are followed by the agent.
  • Sanitization: Database records are processed into objects, but the content of these records is not sanitized or escaped for potential natural language instructions before being passed to the agent context.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 14, 2026, 02:42 PM
Security Audit — agent-trust-hub — byted-viking-aisearch-database