The Agent Skills Directory

[COMMAND_EXECUTION]: The testing script scripts/test_risk_scripts.py uses subprocess.run to execute the bundled utility scripts for verification. It uses the sys.executable path to run the local Python interpreter and does not use shell mode, which is a standard and safe practice for unit tests.
[DATA_EXPOSURE]: The profiling tools scripts/risk_preflight.py and scripts/temporal_split_audit.py are strictly read-only. They process user-provided CSV or JSONL files locally to generate statistical summaries without accessing sensitive system files or making network requests.
[REMOTE_CODE_EXECUTION]: The skill is self-contained and explicitly states that no external dependencies are required. There are no patterns of remote script downloading, package installation, or dynamic code evaluation.
[PROMPT_INJECTION]: The instructions in SKILL.md are well-structured and provide clear methodological guidance and safety boundaries. There are no attempts to override agent behavior, extract system prompts, or bypass safety filters.

actuarial-risk-modeling