analytics
Pass
Audited by Gen Agent Trust Hub on Jul 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions direct the agent to execute shell-based database CLI tools, specifically psql, bq, mysql, and sqlite3. While intended for data analysis, this capability allows the agent to run database commands through the terminal based on generated SQL.
- [PROMPT_INJECTION]: The natural-language-to-SQL feature creates a surface for indirect prompt injection. Maliciously crafted user input or data retrieved from a database could potentially influence the structure or intent of the generated SQL. • Ingestion points: User input strings provided to the /analytics command for translation to SQL. • Boundary markers: The skill includes a 'Show before run' safety rule, requiring the agent to display the generated SQL for user approval before execution. • Capability inventory: Execution of psql, bq, mysql, and sqlite3 CLI tools. • Sanitization: The security model relies on human review of the generated output rather than automated programmatic sanitization.
Audit Metadata