backtest-debugger

What problem does it solve?

Backtests that fail or produce poor performance are hard to debug without systematically inspecting both the outputs and the underlying signal logic; this Skill identifies the root cause and guides targeted repairs.

Core Features & Use Cases

• Artifact-driven diagnosis: Inspects artifacts/metrics.csv, artifacts/equity.csv, and artifacts/trades.csv to classify failures such as runtime errors, abnormal results, and suspicious trading behavior.
• Code-focused root-cause analysis: Reviews code/signal_engine.py and config.json to pinpoint issues like data mapping mismatches, sizing logic bugs, missing exit handling, or dependency-related runtime errors.
• Verification with guardrails: Applies a fix via edit_file, reruns the backtest, and validates outcomes by re-checking updated metrics.csv (including hard-gate checks like non-empty artifacts, NaN-free equity, and exit_code == 0).

Quick Start

Use the backtest-debugger skill to diagnose why your backtest failed or underperformed by inspecting the generated artifacts and updating the signal engine code until the rerun passes validation.