agency-agents-ai-specialists

A collection of specialized AI agent personalities for Claude Code, Cursor, Aider, Windsurf, and other AI coding tools — covering engineering, design, market...

Security Scan

The skill is an instruction-only collection of persona files and its instructions are consistent with that purpose, but it directs you to clone and run scripts from a third‑party GitHub repo — review before running.

Purpose & Capability

Name/description describe a collection of AI agent personalities for Claude Code, Cursor, Aider, Windsurf, Copilot, etc. The SKILL.md explains how to install persona files into those tools' config locations and how to pass files as system prompts. The requested actions (copying/symlinking agent files into tool-specific directories and optionally running conversion/installation scripts) are consistent with the stated purpose.

Instruction Scope

The instructions stay on‑topic: cloning a repository, copying/symlinking persona files into tool-specific directories, and passing persona files as system prompts. They do request reading/writing files under user home and project roots (e.g., ~/.claude/agents, .cursor/rules, .github/copilot-instructions.md), which is expected for installing agent prompts. They do not instruct the agent to read unrelated system secrets or environment variables. However, the installer steps tell the user to run scripts (./scripts/convert.sh and ./scripts/install.sh) from the cloned repo — these scripts could perform arbitrary actions, so inspect them before execution.

Install Mechanism

There is no formal install spec in the registry, but SKILL.md instructs a git clone of https://github.com/msitarzewski/agency-agents.git and running bundled scripts. That is effectively a download-and-run pattern from a third‑party GitHub repo. While GitHub is a normal host, running unknown repo scripts (extract/write/execute) is higher risk than instruction-only operations because those scripts can modify files, run network calls, or execute arbitrary code. The install writes into user home and project directories and suggests modifying repository files (.github/copilot-instructions.md), which increases blast radius if you commit changes.

Credentials

The skill declares no required environment variables, no credentials, and no config paths outside of the expected agent/tool locations. The SKILL.md itself does not request secrets or unrelated credentials. That is proportionate to a prompts/personas collection.

Persistence & Privilege

The skill does not request always:true and does not declare extra privileges. Installation involves writing files into user tool directories and project roots, which is expected for installing persona files. Be aware these file writes may be persistent and can affect projects or repos (e.g., committing .github/copilot-instructions.md).

Assessment

This skill appears to be what it says: a collection of persona files and installation instructions. The main risk is that SKILL.md tells you to git clone a third‑party GitHub repository and run scripts from it. Before installing: 1) Clone the repo into a temporary directory (not root or your production environment). 2) Inspect ./scripts/install.sh and ./scripts/convert.sh for network calls, deletion commands, sudo usage, or unexpected behavior. 3) Prefer copying only the specific Markdown persona files you want rather than running the installer; avoid automatically overwriting tool configs you care about. 4) Run any installer scripts inside a sandbox/container or as a non‑privileged user. 5) Back up files (e.g., .github/copilot-instructions.md) before overwriting and avoid committing automated changes until you've reviewed them. 6) If you don't recognize the repo owner, treat it as untrusted until provenance and intent are verified.

🎭 Agency Agents — AI Specialist Personalities

Skill by ara.so — Daily 2026 Skills collection.

A curated collection of 50+ specialized AI agent personalities for Claude Code, Cursor, Aider, Windsurf, Copilot, and more. Each agent has deep domain expertise, a distinct personality, defined workflows, and measurable deliverables — covering engineering, design, marketing, sales, paid media, and beyond.

---

Installation

Prerequisites

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

Claude Code (Recommended)

# Copy all agents to Claude's agents directory
cp -r agency-agents/* ~/.claude/agents/

# Or symlink for auto-updates
ln -s /path/to/agency-agents ~/.claude/agents/agency

Then in any Claude Code session:

Hey Claude, activate Frontend Developer mode and help me build a React component

All Other Tools (Interactive Installer)

# Step 1: Generate integration files for all supported tools
./scripts/convert.sh

# Step 2: Auto-detect installed tools and install interactively
./scripts/install.sh

# Or target a specific tool
./scripts/install.sh --tool cursor
./scripts/install.sh --tool copilot
./scripts/install.sh --tool aider
./scripts/install.sh --tool windsurf

Manual per Tool

Tool	Install path
Claude Code	~/.claude/agents/
Cursor	.cursor/rules/ in project root
Copilot	.github/copilot-instructions.md
Aider	.aider.conf.yml or pass via --system-prompt
Windsurf	.windsurf/rules/ in project root

---

Agent Roster

Engineering Division

engineering/engineering-frontend-developer.md       React/Vue/Angular, UI, Core Web Vitals
engineering/engineering-backend-architect.md        API design, databases, scalability
engineering/engineering-mobile-app-builder.md       iOS/Android, React Native, Flutter
engineering/engineering-ai-engineer.md              ML models, AI integration, data pipelines
engineering/engineering-devops-automator.md         CI/CD, infra automation, cloud ops
engineering/engineering-rapid-prototyper.md         MVPs, POCs, hackathon speed
engineering/engineering-senior-developer.md         Laravel/Livewire, advanced patterns
engineering/engineering-security-engineer.md        Threat modeling, secure code review
engineering/engineering-code-reviewer.md            PR reviews, code quality gates
engineering/engineering-database-optimizer.md       PostgreSQL/MySQL tuning, slow queries
engineering/engineering-git-workflow-master.md      Branching, conventional commits
engineering/engineering-software-architect.md       System design, DDD, trade-off analysis
engineering/engineering-sre.md                      SLOs, error budgets, chaos engineering
engineering/engineering-incident-response-commander.md  Incident management, post-mortems
engineering/engineering-technical-writer.md         Developer docs, API reference
engineering/engineering-data-engineer.md            Data pipelines, lakehouse, ETL/ELT

Design Division

design/design-ui-designer.md                        Visual design, component libraries
design/design-ux-researcher.md                      User testing, behavior analysis
design/design-ux-architect.md                       CSS systems, technical UX
design/design-brand-guardian.md                     Brand identity and consistency
design/design-whimsy-injector.md                    Micro-interactions, delight, Easter eggs
design/design-image-prompt-engineer.md              Midjourney/DALL-E/SD prompts
design/design-inclusive-visuals-specialist.md       Representation, bias mitigation

Marketing, Sales & Paid Media

marketing/marketing-growth-hacker.md
marketing/marketing-content-creator.md
paid-media/paid-media-ppc-strategist.md
paid-media/paid-media-creative-strategist.md
sales/sales-outbound-strategist.md
sales/sales-deal-strategist.md
sales/sales-discovery-coach.md

---

Using Agents in Claude Code

Activating a Single Agent

# In Claude Code chat:
Activate the Backend Architect agent and help me design a REST API for a multi-tenant SaaS app.

Using Multiple Agents in Sequence

# First, design the system
Activate the Software Architect agent. Design the domain model for an e-commerce platform.

# Then implement
Now activate the Senior Developer agent and implement the Order aggregate in Laravel.

# Then review
Activate the Code Reviewer agent and review the implementation above.

Referencing an Agent File Directly

# Pass an agent as a system prompt in Claude CLI
claude --system-prompt "$(cat ~/.claude/agents/engineering-frontend-developer.md)" \
  "Build a responsive product card component in React with Tailwind CSS"

---

Using Agents in Cursor

After running ./scripts/install.sh --tool cursor, agent rules land in .cursor/rules/. Reference them in chat:

@engineering-frontend-developer Build a data table component with sorting and pagination.

Or set a default rule in .cursor/rules/default.mdc:

---
alwaysApply: true
---

You are operating as the Senior Developer agent from The Agency.
Refer to .cursor/rules/engineering-senior-developer.md for your full persona and workflows.

---

Using Agents with Aider

# Use a single agent as the system prompt
aider --system-prompt "$(cat agency-agents/engineering/engineering-security-engineer.md)"

# Or reference in .aider.conf.yml
echo "system-prompt: agency-agents/engineering/engineering-devops-automator.md" >> .aider.conf.yml

---

Using Agents in Windsurf

./scripts/install.sh --tool windsurf
# Agents are written to .windsurf/rules/

Activate in chat:

Use the UX Architect agent rules from .windsurf/rules/ to audit my CSS architecture.

---

Real Workflow Examples

Full-Stack Feature with Multiple Agents

# 1. Architecture phase
cat > task.md << 'EOF'
I need to add real-time notifications to my Node.js + React app.
Users should see in-app alerts and optionally receive email digests.
EOF

# Invoke Software Architect
claude --system-prompt "$(cat ~/.claude/agents/engineering-software-architect.md)" < task.md

# 2. Backend implementation
claude --system-prompt "$(cat ~/.claude/agents/engineering-backend-architect.md)" \
  "Implement the notification service based on the architecture above using PostgreSQL LISTEN/NOTIFY and Socket.io"

# 3. Frontend implementation
claude --system-prompt "$(cat ~/.claude/agents/engineering-frontend-developer.md)" \
  "Build the React notification bell component that connects to the Socket.io feed"

# 4. Security review
claude --system-prompt "$(cat ~/.claude/agents/engineering-security-engineer.md)" \
  "Review the notification system implementation for security issues"

Code Review Workflow

# Generate a diff and pipe to the Code Reviewer agent
git diff main..feature/payment-integration | \
  claude --system-prompt "$(cat ~/.claude/agents/engineering-code-reviewer.md)" \
  "Review this PR diff. Focus on security, correctness, and maintainability."

Database Optimization

# Paste slow query log and activate Database Optimizer
claude --system-prompt "$(cat ~/.claude/agents/engineering-database-optimizer.md)" << 'EOF'
Here is a slow query from our PostgreSQL logs (avg 4200ms):

SELECT u.*, p.*, o.*
FROM users u
LEFT JOIN profiles p ON p.user_id = u.id
LEFT JOIN orders o ON o.user_id = u.id
WHERE u.created_at > NOW() - INTERVAL '30 days'
ORDER BY o.created_at DESC;

Table sizes: users=2M rows, orders=18M rows. No indexes on created_at columns.
EOF

Incident Response

# Structured incident kick-off
claude --system-prompt "$(cat ~/.claude/agents/engineering-incident-response-commander.md)" << 'EOF'
SEV-1 INCIDENT: Payment processing returning 503 errors since 14:32 UTC.
Error rate: 94%. Affected: checkout, subscription renewals.
Recent deploys: payment-service v2.4.1 at 14:15 UTC.
EOF

---

Creating Custom Agents

Agent files follow a consistent markdown structure:

# 🎯 Agent Name

## Identity
You are [Name], [role] at The Agency...

## Core Mission
[What this agent optimizes for]

## Personality & Communication Style
- [Trait 1]
- [Trait 2]

## Workflows

### [Workflow Name]
1. [Step 1]
2. [Step 2]

## Deliverables
- [Concrete output 1]
- [Concrete output 2]

## Success Metrics
- [Measurable outcome]

Save custom agents to agency-agents/custom/ and re-run ./scripts/convert.sh to generate tool integrations.

---

Contributing New Agents

# Fork and clone
git clone https://github.com/YOUR_USERNAME/agency-agents.git

# Create your agent in the appropriate division
cp engineering/engineering-senior-developer.md \
   engineering/engineering-YOUR-SPECIALTY.md

# Edit the file, then test it
claude --system-prompt "$(cat engineering/engineering-YOUR-SPECIALTY.md)" \
  "Give me a sample deliverable to demonstrate your capabilities"

# Submit a PR
git checkout -b agent/your-specialty
git add engineering/engineering-YOUR-SPECIALTY.md
git commit -m "feat: add Your Specialty agent"
git push origin agent/your-specialty

---

Troubleshooting

Agents not found in Claude Code

ls ~/.claude/agents/
# If empty, re-run:
cp -r /path/to/agency-agents/* ~/.claude/agents/

convert.sh fails with permission error

chmod +x scripts/convert.sh scripts/install.sh
./scripts/convert.sh

Cursor not picking up agent rules

# Rules must be in project root .cursor/rules/
ls .cursor/rules/
# Re-run installer targeting cursor
./scripts/install.sh --tool cursor

Agent personality not activating

• Be explicit: "Activate the Frontend Developer agent" rather than just referencing the topic
• Paste the agent file contents directly into the system prompt if tool integration isn't working
• For Claude Code, confirm agents directory: claude config get agentsDir

Agent conflicts when using multiple

• Activate one agent per conversation session
• For multi-agent workflows, use separate sessions or Claude Code's subagent feature
• Sequence agents explicitly: architect → implement → review

---

Project Structure

agency-agents/
├── engineering/          # 23 engineering specialist agents
├── design/               # 8 design specialist agents
├── marketing/            # Marketing and growth agents
├── sales/                # 8 sales specialist agents
├── paid-media/           # 7 paid media specialist agents
├── scripts/
│   ├── convert.sh        # Generate tool-specific integration files
│   └── install.sh        # Interactive installer (auto-detects tools)
└── README.md

---

Key Facts

• License: MIT
• 51,000+ stars — battle-tested by a large community
• No API keys required — agents are prompt files, not services
• Tool-agnostic — works with any LLM tool that accepts system prompts
• Extensible — add custom agents following the same markdown pattern
• PRs welcome — the roster grows through community contributions