explaingit

msitarzewski/agency-agents

99,982ShellAudience · vibe coderComplexity · 1/5MaintainedLicenseSetup · easy

TLDR

Pre-written AI agent personas for coding assistants. Copy agent definitions into Claude, Cursor, or other tools to give your AI a specialized role like frontend developer, security engineer, or code reviewer.

Mindmap

mindmap
  root((agency-agents))
    What it does
      Pre-built personas
      Role definitions
      Ready-made experts
    How to use
      Copy to AI tools
      Activate in session
      Install via scripts
    Supported tools
      Claude Code
      Cursor
      GitHub Copilot
      Aider
    Agent types
      Frontend Developer
      Backend Architect
      Security Engineer
      Database Optimizer
    Use cases
      Skip prompt writing
      Assemble AI teams
      Study prompt structure

Things people build with this

USE CASE 1

Activate a specialized AI persona (e.g., 'Frontend Developer mode') in your coding assistant without writing a custom prompt.

USE CASE 2

Build a multi-agent workflow by installing different personas and switching between them for different project tasks.

USE CASE 3

Study how detailed role prompts are structured to write better custom personas for your own AI tools.

Tech stack

ShellMarkdown

Getting it running

Difficulty · easy Time to first run · 5min
Use freely for any purpose, including commercial use, as long as you keep the copyright notice.

In plain English

Agency-agents is a collection of pre-written persona files for AI coding assistants. The author calls it a complete AI agency at your fingertips: each agent is a markdown file describing a specialised role, a frontend developer, backend architect, mobile builder, security engineer, technical writer, code reviewer, database optimiser, and many others organised into divisions like Engineering. According to the README each agent has its own identity and personality, a core mission, example workflows, expected technical deliverables, and success metrics, so it can be used as a ready-made expert rather than a blank prompt template. In practice, you do not run this repository as software. Instead, you copy or install the agent definitions into one of the supported AI coding tools so the assistant takes on that role when activated. The README ships shell scripts that install the agents to Claude Code by default and can also generate integration files for other tools the project lists as supported, including GitHub Copilot, Antigravity, Gemini CLI, OpenCode, OpenClaw, Cursor, Aider, Windsurf, and Kimi Code. After installation you activate an agent inside a session, for example by saying "activate Frontend Developer mode", and the assistant follows that persona's instructions. Someone might use it to skip writing prompts from scratch when starting a new task, to assemble a small team of focused assistants for different parts of a project, or to study how detailed role prompts are structured. The repository is shell-script-driven and is released under the MIT licence.

Copy-paste prompts

Prompt 1
Install the agency-agents personas into my Claude Code setup and show me how to activate the Backend Architect agent.
Prompt 2
Generate the integration files for Cursor from the agency-agents repository so I can use these personas in my editor.
Prompt 3
Show me the Frontend Developer agent definition and explain what makes it effective as a coding assistant persona.
Prompt 4
How do I create a custom agent persona following the structure used in agency-agents for my own specialized role?
Open on GitHub → Explain another repo

Generated 2026-05-18 · Model: sonnet-4-6 · Verify against the repo before relying on details.