agency-agents: A Full AI Specialist Team for Your Development Tools

msitarzewski/agency-agents packages professional roles as reusable Agent definition files so AI coding tools can switch expert perspectives by task—not rely on one generic assistant.

At a Glance

msitarzewski/agency-agents is an Agent role library for AI coding tools and multi-agent workflows. It is not a single app or a loose prompt collection—it is a set of AI specialist definitions organized by professional team.

These Agents span engineering, design, marketing, sales, product, project management, testing, security, GIS, game development, finance, academic research, and more. They install into Claude Code, Cursor, Codex, Gemini CLI, OpenCode, Aider, Windsurf, Kimi Code, Osaurus, Hermes, and similar tools.

What It Is

Each Agent file typically includes:

• Identity and persona.
• Working style.
• Task boundaries.
• Deliverable formats.
• Success criteria.
• Communication style.

Think of it as an AI expert library for switching professional perspectives by task—not all-in-one software that automatically does everything.

Who It’s For

• Developers who live in Claude Code, Codex, Cursor, etc. and want fast role switching.
• Teams building multi-agent workflows who need ready-made division of labor, boundaries, and deliverable standards.
• Anyone learning how to write high-quality Agent definition files—persona, process, constraints, and output format.

Using it well means picking the right role, providing clear context, and folding Agent output into code review, testing, compliance, and release processes.

Installation

The project recommends the desktop app. The official Agency Agents App supports macOS, Linux, and Windows: browse the full roster and one-click install into Claude Code, Cursor, Codex, Gemini, Osaurus, etc., with updates handled for you.

On macOS via Homebrew:

brew install --cask msitarzewski/agency-agents/agency-agents

Command-line setup:

git clone https://github.com/msitarzewski/agency-agents.git
cd agency-agents

Install into Claude Code:

./scripts/install.sh --tool claude-code

Copy a single division, e.g. engineering:

cp engineering/*.md ~/.claude/agents/

Before installing to other tools, generate integration files:

./scripts/convert.sh

Interactive install:

./scripts/install.sh

Install to a specific tool:

./scripts/install.sh --tool cursor
./scripts/install.sh --tool codex
./scripts/install.sh --tool gemini-cli
./scripts/install.sh --tool opencode
./scripts/install.sh --tool aider
./scripts/install.sh --tool windsurf
./scripts/install.sh --tool kimi
./scripts/install.sh --tool osaurus
./scripts/install.sh --tool hermes

Select by division or agent:

./scripts/install.sh --tool claude-code --division engineering,security
./scripts/install.sh --tool cursor --agent frontend-developer,ui-designer
./scripts/install.sh --list teams
./scripts/install.sh --tool opencode --division engineering --dry-run

OpenCode currently registers about 119 Agents at runtime; extras may be silently dropped. Prefer --division subsets when installing for OpenCode.

How to Use It

After installation, switch roles directly in your AI coding tool:

Activate Frontend Developer mode and help me build a React component.

More practical prompts include task, context, and expected deliverables:

Activate Backend Architect.
Review this API design, identify scaling risks, and propose a database schema.
Keep the output as decisions, trade-offs, and next implementation steps.

For complex projects, combine multiple Agents instead of one general assistant. For example:

• Product Manager: PRD and acceptance criteria.
• Software Architect: system boundaries.
• Backend Architect: API and data design.
• Frontend Developer: UI implementation.
• API Tester and Reality Checker: validation.
• Technical Writer: documentation.

Each role owns one clear lens—quality control gets easier.

Agent Categories

A compressed index of the library’s coverage:

Engineering

Includes Frontend Developer, Backend Architect, Mobile App Builder, AI Engineer, DevOps Automator, Software Architect, SRE, Code Reviewer, Database Optimizer, Prompt Engineer, Multi-Agent Systems Architect, and more.

Covers full-stack development, architecture, AI integration, CI/CD, reliability, code review, database tuning, prompt design, and multi-agent governance.

Design

Includes UI Designer, UX Researcher, UX Architect, Brand Guardian, Visual Storyteller, Image Prompt Engineer, Persona Walkthrough Specialist, and more.

Covers UI design, user research, brand consistency, visual storytelling, AI image prompts, and persona walkthroughs.

Paid Media and Sales

Paid Media covers ad accounts, search terms, tracking and measurement, creative strategy, display, and social ads.

Sales covers outbound strategy, discovery calls, deal strategy, pre-sales engineering, RFP proposals, pipeline analysis, expansion, and sales coaching.

Useful for growth, B2B sales, and paid media teams building specialized AI roles.

Marketing

Includes Growth Hacker, Content Creator, SEO Specialist, LinkedIn Content Creator, Xiaohongshu Specialist, Douyin Strategist, WeChat Official Account Manager, Podcast Strategist, AI Citation Strategist, AEO Foundations Architect, and more.

Spans content, social, SEO, short video, private traffic, e-commerce, AI search visibility, and multi-platform publishing.

Product and Project Management

Product: Product Manager, Sprint Prioritizer, Trend Researcher, Feedback Synthesizer, Behavioral Nudge Engine.

Project Management: Senior Project Manager, Project Shepherd, Experiment Tracker, Meeting Notes Specialist, Jira Workflow Steward.

Covers discovery, roadmaps, feedback, delivery, experiments, and meeting notes.

Testing

Includes Evidence Collector, Reality Checker, Test Results Analyzer, Performance Benchmarker, API Tester, Accessibility Auditor.

Covers UI evidence, release gates, test analysis, performance testing, API testing, and accessibility audits.

Security

Includes Security Architect, Application Security Engineer, Penetration Tester, Cloud Security Architect, Incident Responder, Threat Detection Engineer, Compliance Auditor, Blockchain Security Auditor.

Covers threat modeling, security architecture, code review, authorized pentesting, cloud security, detection engineering, compliance prep, and smart contract audits.

Support, Finance, and Specialized

Support: customer service, analytics reports, finance tracking, infrastructure maintenance, legal compliance, executive summaries.

Finance: bookkeeping, financial analysis, FP&A, investment research, tax strategy.

Specialized: MCP Builder, Document Generator, Automation Governance Architect, Developer Advocate, Data Privacy Officer, Chief of Staff, Customer Success Manager, Workflow Architect, Salesforce Architect, and more.

Spatial Computing, Game Development, Academic, and GIS

Spatial Computing: XR, visionOS, Metal, WebXR, terminal integration.

Game Development: Unity, Unreal, Godot, Roblox, level design, narrative, audio, shaders, multiplayer architecture, editor tooling.

Academic: anthropology, geography, history, narratology, psychology.

GIS: cartography, spatial data engineering, ArcPy, GeoAI, BIM/GIS, 3D scenes, drone mapping, Web GIS.

Real-World Combinations

Staff by project phase—not one role for everything.

Startup MVP

Combine:

• Frontend Developer
• Backend Architect
• Growth Hacker
• Rapid Prototyper
• Reality Checker

Covers implementation, architecture, acquisition paths, rapid validation, and pre-release quality checks.

Marketing Campaign

Combine:

• Content Creator
• Twitter Engager
• Instagram Curator
• Reddit Community Builder
• Analytics Reporter

Each platform gets the right tone, format, and execution focus.

Enterprise Feature Delivery

Combine:

• Senior Project Manager
• Senior Developer
• UI Designer
• Experiment Tracker
• Evidence Collector
• Reality Checker

Fits teams that need scope management, implementation, design, experiments, and acceptance loops.

Security or Infrastructure Work

Don’t ask non-security roles to “take a quick look.” A better sequence:

1. Security Architect for threat modeling.
2. Application Security Engineer or Penetration Tester for code and attack surface review.
3. Incident Responder for runbooks and response flows.

Recommendations

1. Install selectively. Too many roles clutter the tool. Start with Engineering, Testing, Security; add Marketing, Paid Media, Sales when needed.
2. Set clear boundaries. If Minimal Change Engineer fixes a bug, say no refactors. If Software Architect evaluates design, don’t ask it to ship code in the same turn.
3. Treat output as advice, not gospel. Run tests on code, verify security findings, and align marketing/compliance guidance with your business and regulations.
4. Study the files as templates. Good Agent definitions include boundaries, workflow, deliverable format, quality bars, and failure handling—not just “you are an expert in X.”

Summary

agency-agents turns professional roles into reusable Agent definition files so AI coding tools can switch expert perspectives instead of relying on one generic assistant.

Use it as an expert role library and Agent definition reference: pick a small set of high-frequency roles, build combinations around real workflows, and validate output through testing, review, and release discipline.