The Agent Skills Directory

[SAFE]: The skill implements a robust audit process for medical model validation studies using local scripts. No malicious intent, obfuscation, or unauthorized access to sensitive data was detected.
[COMMAND_EXECUTION]: The skill uses the Bash tool to execute a local Python script scripts/check_split_leakage.py. This script is a utility for scientific integrity that performs set arithmetic on split tables. It relies only on the Python standard library and does not perform any network operations or subprocess calls.
[PROMPT_INJECTION]: The skill includes instructions to process user-provided CSV files for model validation. While this involves ingesting external data, the process is well-guarded through the use of deterministic scripts.
Ingestion points: Data enters the process via the split_assignment.csv file specified in Phase 2.
Boundary markers: The skill instructs the agent to rely on the deterministic output of the script rather than interpreting data through natural language reasoning, which effectively isolates the agent context from the raw data content.
Capability inventory: The skill uses standard Bash, Read, Write, and Edit tools to perform its auditing tasks.
Sanitization: The check_split_leakage.py script parses the input CSV using standard libraries and enforces a specific schema (patient IDs and partitions), providing structural validation of the input data before the results are presented to the agent.

model-validation