bigtable-basics
Pass
Audited by Gen Agent Trust Hub on Sep 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- Command Injection Surface: The skill includes templates for gcloud and cbt CLI commands that use placeholders for user-defined values such as table names, row keys, and search queries. This represents a potential surface for indirect prompt injection if user inputs are not properly handled before command generation.\n
- Ingestion points: Multiple reference files (cli_data_access.md, dataplex.md, infrastructure_management.md) use placeholders like {table}, {row_key}, and {search_term}.\n
- Boundary markers: The skill contains a specific safety rule in SKILL.md requiring the agent to obtain explicit user confirmation before executing commands that modify the database.\n
- Capability inventory: The skill utilizes shell command execution via official Google Cloud administration tools.\n
- Sanitization: The reference materials focus on command structure and do not provide explicit input validation or escaping logic for the interpolated variables.\n- External API Interaction: The skill demonstrates how to search Bigtable assets using the Dataplex API via curl. This interaction targets official Google Cloud endpoints and follows standard authentication practices using gcloud application-default tokens.\n- Configuration Management: The skill provides instructions for configuring the cbt tool by writing environment-based settings to ~/.cbtrc. This is a standard and documented method for managing tool-specific configuration.
Audit Metadata