Orchestration
Multi-agent workflows, parallel fan-out and clean delegation to subagents.
AgenticLevel by Bailder
Anyone runs an agent. Few conduct a fleet.
One number for how well you orchestrate, verify and let a fleet of AI agents run — 0 to 100, a grade, and your archetype. Pulled straight from your own transcripts.
gh skill install bailder/agenticlevel
Runs locally. Reads only structural metadata from ~/.claude/projects — never your code or prompts.
The conductors
Opt-in, ranked by mastery — not by how much they type.
No conductors on the board yet. Be the first.
Full leaderboard →What it measures
Six pillars of conducting
Weighted for mastery. Volume is deliberately not one of them.
Verification
Tests, builds and re-reads after edits — proof, before shipping.
Tooling
Breadth of tools, advanced tool-search and skills, ecosystem reach.
Delivery
Pull requests, magnitude of change and orchestration efficiency.
Craft & Discipline
Few compactions, API errors, denials or interruptions. Clean runs.
Autonomy
Scheduled loops, AFK leverage and consistency across the month.
Why not the volume score
Same top grade. Different proof.
On a volume-weighted score, most of the number is just how many messages you sent. The same spam that scores 92 there earns a D here.
Volume score
92 — for spamming the agent
AgenticLevel
D — same spam, no mastery
We don't reward keystrokes. We reward conducting.
Get your score