The Agent Skills Directory

[COMMAND_EXECUTION]: Orchestrates data movement by executing shell commands via subprocess.run, specifically utilizing the AWS CLI (aws s3 cp) to transfer datasets between Amazon S3 buckets and the local file system.
[DYNAMIC_EXECUTION]: Implements a workflow that dynamically generates Python transformation functions and execution scripts. These are saved to disk and executed using the python3 interpreter to validate the transformation logic against sample data records.
[INDIRECT_PROMPT_INJECTION]: Ingests and processes untrusted data from user-provided datasets. The skill lacks explicit sanitization or boundary markers when presenting sample records in the agent's context, which could allow malicious instructions embedded in a dataset to influence the agent's behavior.
Ingestion points: Reads sample records from local or S3-hosted datasets in SKILL.md Step 5 and Step 11.
Boundary markers: None identified in the instructions for displaying sample records.
Capability inventory: Subprocess shell execution, file system write access, and dynamic Python code execution.
Sanitization: No validation or filtering of dataset content is performed before processing or display.
[EXTERNAL_DOWNLOADS]: Fetches the latest evaluation dataset schema definitions from official AWS documentation at docs.aws.amazon.com.

dataset-transformation