lookahead-safety

What problem does it solve?

This Skill prevents look-ahead bias in quant research by ensuring you only use data that was actually published and therefore knowable at the time you claim to know it.

Core Features & Use Cases

• Enforces known-date semantics: treats filing_date / acceptedDate (or equivalent publication timestamps) as the availability boundary, not period_end or other period-of-report fields.
• Hardens SEC/XBRL workflows: avoids common SEC-specific traps like XBRL period_end leakage, filing amendments (10-K/A superseding), and unknown publication windows.
• Improves backtest integrity: requires query_date-bounded historical reads so future records cannot contaminate past calculations, including explicit “future-data leak” testing guidance.

Quick Start

Use the lookahead-safety skill when you ask an AI to “compute metric X as of date D from SEC/XBRL/backtest data” so it filters out any records with filing_date after D and clearly states the admissibility rule.