Most AI-agent collections are just a folder of reworded system prompts. This one is laid out like a company: 232 named specialists sorted into 16 divisions, each with a job description, a workflow, and the deliverables it's expected to produce. The underlying bet is that orchestrating AI work gets easier when agents map to roles a real org would actually staff, not to vague "assistant" personas.
What Sets It Apart
- Roles, not prompts. Each agent carries a defined scope, communication style, process, and measurable success metrics — so you reason about which agent to invoke instead of pasting a prompt and hoping.
- Org-chart structure. The 16 divisions (engineering, design, marketing, sales, product, security, finance, game development, and more) make a 232-agent library navigable; you pick a department the way you'd pick a hire.
- Built around outputs. Agents are written around the artifacts they produce — real code, documents, processes — rather than open-ended chat.
- Community origin. It grew out of a Reddit thread on agent specialization, which shows in the breadth: niche roles like a "reality checker" sit beside the obvious engineering ones.
Who It's For
Great fit if you're assembling a multi-agent workflow — Claude Code subagents, an orchestration framework — and want a ready-made roster instead of authoring every persona yourself. Look elsewhere if you need one carefully tuned agent for a single narrow task; 232 personas is breadth you won't use. Treat these as editable definitions to adapt to your stack, not a runtime or an installable tool.
