A self-improving evaluation system that scores orchestration quality across five weighted metrics. Like observing stars through a telescope, we measure the brilliance of AI orchestration.
Drop your .md, .txt, or .jsonl file. Parser extracts orchestration signals: skills invoked, agents spawned, decision points, error recoveries, learning signals.
Scoring engine evaluates across five weighted metrics. Each metric scored 0-100, combined into composite score with mathematical precision.
Session appears on public leaderboard with full breakdown. Filter by score range, project type, or skill category.
Top sessions analyzed for winning patterns. System improves its own scoring weights through compound learning loops.
This system was built in a single AI session and will evaluate itself. The orchestration patterns used to build SessionStellar become the first entries in its own star chart—a recursive validation of its scoring algorithm.