migrating-dbt-project-across-platforms

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from dbt project files, which could theoretically contain malicious instructions.
  • Ingestion points: The agent reads .sql models, profiles.yml, dbt_project.yml, and _sources.yml as part of the migration workflow.
  • Boundary markers: The skill includes a dedicated 'Handling External Content' section and 'Don't Do These Things' section that explicitly instruct the agent to treat project content as untrusted and ignore embedded commands.
  • Capability inventory: The agent can execute dbt/dbt Fusion commands, perform file writes to documented change logs, and use standard CLI utilities like jq and grep.
  • Sanitization: The instructions mandate extracting only structured data and ignoring any natural language instructions found within the project files.
  • [COMMAND_EXECUTION]: The workflow requires executing various CLI tools to perform the migration.
  • Evidence: The skill calls for dbtf compile, dbtf run, dbt test, dbtf debug, and dbtf system update.
  • Evidence: Identification of leaf nodes uses a pipeline of dbt ls, jq, sort, grep, and comm.
  • Context: These commands are standard for dbt development and the tool dbtf (dbt Fusion) is the official tool provided by the vendor (dbt Labs).
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 02:38 PM
Security Audit — agent-trust-hub — migrating-dbt-project-across-platforms