A 179-Person AI Workforce in a Node Package

What it does

This repo is a corporate directory translated for LLMs: 179 pre-built agent personas spanning 13 departments—from UX researchers to Solidity engineers—packaged as a CLI tool. You install roles (built-in, local, or from any GitHub repo) and activate them by typing /<agent-name> inside a compatible chat harness.

The interesting bit

Instead of building autonomous tooling, the project treats “agents” as reusable prompt costumes. The underlying Hindsight engine presumably handles memory and learning, while this repository focuses on curating the org chart.

Key highlights

• 179 agents across 13 departments, each distilled to a markdown persona with a specific “vibe”
• CLI supports installing from the built-in roster, local directories, or arbitrary GitHub repo paths
• Harness-agnostic-ish: supports Claude, claude-code, openclaw, nemoclaw, and hermes
• Runs offline once installed; personas are local markdown definitions
• MIT licensed and distributed as a Node package

Caveats

• The README catalogs personas extensively but never explains how “self-learning” actually works; the Hindsight dependency is referenced but not detailed
• Integration with Claude Chat requires manual skill upload and API host allowlisting, adding friction
• With 179 markdown persona files and no visible runtime logic in the README, the heavy lifting appears delegated to the harness and Hindsight

Verdict

Grab it if your team wastes hours rewriting the same “act as a senior DevOps engineer” prompts. Skip it if you need autonomous agents with native tool use, memory, or orchestration—this is a costume closet, not a robot factory.