Confidence Scorer Skill

Confidence scoring for technology stack identification works by weighting multi-source evidence such as source diversity, signal strength, and cross-source agreement. It then applies conflict penalties to assign High, Medium, or Low labels to each correlated technology finding.

When multi-source evidence disagrees during technology stack identification, this skill applies conflict penalties to penalize the agreement score, resulting in lowered confidence labels and flagging specific gaps where additional validation signals are needed.

The best way to resolve uncertainty from incomplete corroboration is to use confidence scoring to convert multi-source evidence signals into consistent High, Medium, or Low labels, producing a prioritized list of confidence gaps for follow-up verification.

Confidence scoring requires a structured correlated_technologies input array containing sources, agreement_score, and corroborating_signals. An optional conflicts array can be included alongside this data to apply conflict penalties and threshold-based confidence mapping.

You identify confidence gaps in a technology stack analysis by processing weighted evidence and conflict impacts. The skill produces an actionable list of confidence gaps that indicates exactly where additional validation signals are needed across the report.