Skill: Hypothesis Engine (Working-Theory Ledger)

What problem does it solve?

It prevents investigation anchoring bias by maintaining an explicit, continuously updated ledger of 2–3 working theories that can be rescored and retired as new evidence arrives.

Core Features & Use Cases

• Append-only hypothesis ledger: Records hypothesis evolution in JSONL at analysis/hypotheses.jsonl without rewriting prior entries.
• Evidence-driven scoring: Uses scripts/hypothesis_score.py to compute confidence mechanically from weighted supportive and contradictory signals.
• Operational-period maintenance: Re-scores active hypotheses at each operational period boundary, retires low-confidence items, and ensures at least one adequately supported hypothesis exists.
• Contradiction surfacing: Detects mutually exclusive high-confidence hypotheses and surfaces contradictions in the COP’s Contradictions table.
• Report-ready reasoning: Produces artifacts that downstream reporting can render into narrative explanations of how conclusions were reached.

Quick Start

Invite the Investigation Section Chief to invoke this skill at case start and then at every operational period boundary to keep hypotheses scored, retired, and updated based on the latest COP findings.