agentskills.codes
VE

verify-regime

>

Install

mkdir -p .claude/skills/verify-regime && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/14641" && unzip -o skill.zip -d .claude/skills/verify-regime && rm skill.zip

Installs to .claude/skills/verify-regime

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

Level 5 verification: regime conditioning tests across VIX regimes, stress event validations (March 2020, February 2018), and A-B data source comparisons. Runs after verify-empirical passes.
190 charsno explicit “when” trigger

About this skill

Verify Regime — Level 5

Command

uv run pytest tests/regime/ -v

Regime definitions

VIX-based thresholds applied to data/computed/regime_labels.csv:

RegimeVIX rangeExpected behavior
Low< 12Small parity deviations, positive VRP
Normal12–20Baseline results hold
Elevated20–30Wider deviations, VRP signal noisy
Crisis> 30Large parity deviations, VRP may flip

What regime tests check

  • Parity deviation by regime: crisis deviation > normal deviation (statistically)
  • Backtest performance by regime: strategy Sharpe higher in normal than crisis
  • VRP signal: positive mean in low/normal, check sign in elevated/crisis
  • Regime transition: no artificial jumps at threshold boundaries

Stress event validations

Specific dates that must show expected behavior:

EventDateExpected
COVID crash2020-03-16VIX > 80, parity deviation > 2%
VIX spike2018-02-05VIX > 37, large call/put spread widening
GFC2008-10-10If in sample: VIX > 70

A-B data source comparison

FRED VIXCLS vs CBOE VIX historical should match within 0.01 vol points. Run: pytest tests/regime/test_source_comparison.py -v

Current status

tests/regime/ and data/computed/regime_labels.csv not yet built. Depends on data-computed skill producing regime labels from VIX thresholds.

Search skills

Search the agent skills registry