duckdb-analytics

Pass

Audited by Gen Agent Trust Hub on May 20, 2026

Risk Level: SAFECREDENTIALS_UNSAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The Python ETL example contains hardcoded credentials ('admin' and 'password') intended for a local MinIO storage endpoint.
  • [COMMAND_EXECUTION]: Code templates utilize string interpolation (f-strings) to construct SQL commands, creating a potential SQL injection vulnerability if parameters like 'endpoint' or 'access_key' are sourced from untrusted user input.
  • [PROMPT_INJECTION]: The skill is designed to ingest data from external Parquet files and S3 buckets, creating a surface for indirect prompt injection. * Ingestion points: External data is read via read_parquet() from S3/MinIO paths in SKILL.md. * Boundary markers: None provided in the instructions to prevent the model from interpreting data content as instructions. * Capability inventory: The skill has access to Bash and Write tools, which could be exploited if malicious data influences the agent's logic. * Sanitization: No input validation or sanitization of the Parquet file content is described.
Audit Metadata
Risk Level
SAFE
Analyzed
May 20, 2026, 02:51 AM
Security Audit — agent-trust-hub — duckdb-analytics