The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script "scripts/run_analysis.py" uses "importlib.util" to dynamically load bundled Python templates (e.g., "diagnostic_accuracy.py"). This is a local file operation used for modular code execution and is implemented with path validation to prevent traversal attacks.
[COMMAND_EXECUTION]: The skill contains internal testing and verification scripts ("tests/test_analysis_run.py", "scripts/analysis_run_challenge/verify.sh") that utilize "subprocess.run" to execute local Python scripts and verify analysis consistency. These operations are restricted to the local environment and intended for quality assurance.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill incorporates a mandatory "Data Privacy Check" in its instructions, directing the agent to prioritize de-identified files and warn users before processing data that may contain Protected Health Information (PHI). This minimizes the risk of accidental PII exposure.
[INDIRECT_PROMPT_INJECTION]: The skill processes external tabular data (CSV/Excel) during the assessment phase. While this represents a data ingestion surface, the skill scopes the agent's interaction to statistical reporting and code generation, reducing the risk of malicious instructions within the data influencing agent behavior.

analyze-stats