migrating-sql-to-dbt
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from legacy SQL files which represents a potential injection surface.
- Ingestion points: The workflow begins by reading local SQL files using the cat command.
- Boundary markers: The skill does not provide specific instructions to the agent to ignore potentially malicious instructions embedded in SQL comments.
- Capability inventory: The agent has access to shell commands and the dbt CLI for file manipulation and model execution.
- Sanitization: There is no explicit sanitization of the SQL input before it is processed into dbt models.
- [COMMAND_EXECUTION]: The skill utilizes standard command-line tools to interact with the project environment.
- It executes grep and find to locate existing models and sources within the project directory.
- It invokes the dbt CLI to build and validate the newly created data models, which is standard behavior for dbt development tasks.
Audit Metadata