Benchmark feature vs develop performance

feature vs develop 브랜치 성능 비교 —

Measuring performance across configuration variants with automated iteration and winner ranking.

• · args

• · maxDeltaPct
• · verdict
• · report

• · agent()

• · Agent may return null; filtering applied

• · 2 structured output schema(s) defined with required fields
• · Includes test files with assertion updates on changes
• · Public GitHub repo: moongci38-oss/forge
• · Canonical meta export with phases and description
• · Explicit phase() calls (2) for workflow transparency