dataset-transformation
Pass
Audited by Gen Agent Trust Hub on Jun 28, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches dataset schema definitions from the official AWS SageMaker documentation to ensure accurate format resolution.
- [COMMAND_EXECUTION]: The skill uses Python's subprocess module to execute AWS CLI commands for data transfers and to run generated transformation scripts.
- [REMOTE_CODE_EXECUTION]: The skill generates Python logic for data transformation and executes it within the user's environment. These operations are core to the skill's functionality and follow strict templates with user oversight.
- [PROMPT_INJECTION]: The skill processes external documentation and datasets, creating a surface for indirect prompt injection. This is mitigated by explicit instructions to ignore embedded commands and validate schemas.
- Ingestion points: Official AWS documentation URLs and user-provided dataset paths.
- Boundary markers: Explicit instructions in Step 3 to treat fetched content as untrusted and ignore any embedded instructions or tool requests.
- Capability inventory: File system access, network interaction via AWS CLI, and execution of local/remote scripts.
- Sanitization: Mandatory schema validation and user review cycles for all generated code.
Audit Metadata