confidence-aggregator

Aggregation combines scores from multiple agents into a single confidence metric. Pass an array of agent scores with their IDs and types to aggregateScores() to compute weighted averages and detect outliers, producing a unified confidence assessment.

Outlier detection uses the interquartile range (IQR) method to identify anomalous scores. The Skill flags scores falling outside typical ranges, helping you exclude erratic evaluations from your final confidence calculation.

Yes, weighted averaging lets you assign importance factors to each agent before aggregation. Agents with higher weights contribute more to the final confidence score, letting you prioritize trusted or specialized evaluators.

Group-by-type analysis segments aggregated scores by agent category—backend, frontend, validator, etc. This reveals whether specific evaluation types consistently diverge, highlighting systematic biases or expertise gaps across your evaluation team.

Combine outlier detection with weighted averaging to resolve conflicts. Remove statistical outliers first, then compute confidence using importance weights for remaining scores, producing a robust consensus that balances majority opinion with expert credibility.