database-designer

Pass

Audited by Gen Agent Trust Hub on Nov 11, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill performs database design analysis using standalone Python scripts that do not require external dependencies or network access. All logic is contained within the provided source code, and scripts are restricted to processing files specified by the user.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect injection as it processes user-provided SQL and JSON schemas. However, the migration_generator.py script specifically hardens output generation against SQL injection, reducing the risk to a safe level.
  • Ingestion points: Untrusted schema data is ingested via schema_analyzer.py (through the --input argument) and migration_generator.py (through the --current and --target arguments).
  • Boundary markers: Absent (the skill relies on input sanitization and strict type validation rather than delimiters).
  • Capability inventory: The skill possesses file system read capabilities for input schemas and file system write capabilities for outputting SQL scripts, Mermaid diagrams, and analysis reports.
  • Sanitization: The skill implements robust sanitization functions including quote_ident for identifier quoting, quote_literal for string literal escaping, and _check_sql_fragment for the validation of SQL expressions (checking for balanced quotes, parentheses, and forbidden characters like ; or --).
  • [DYNAMIC_EXECUTION]: Test scripts within the repository (test_migration_generator.py, test_migration_integration.py) use subprocess.run to execute the skill's own CLI and sqlite3 to validate migration logic in memory. These are standard development and quality assurance practices and do not involve the execution of untrusted external code.
  • [COMMAND_EXECUTION]: The documentation includes instructions for setting file execution permissions (chmod +x), which is standard procedure for localized command-line utilities and does not represent a privilege escalation vector.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 11, 2026, 10:52 AM