👥 Agent-Skins

18,720+ Production-Ready AI Agent Prompts, Tool Schemas & Framework Configs

Stop spending weeks debugging prompts. Pick a role, pick your industry, pick your framework — and ship.

---

What is this?

Every file in this repository is a complete, ready-to-use AI agent configuration. Each folder contains:

• ✅ Framework configurations: Copy-pasteable boilerplate for CrewAI, LangGraph, AutoGen, or Vercel AI SDK.
• ✅ Universal raw prompts (raw_prompt.md): A framework-agnostic system prompt with 2,100+ lines of production-grade depth for ChatGPT, Claude, Gemini, Ollama, or any API.
• ✅ Multi-Point Audit Matrix (AUD-01 to AUD-05): Adherence checklists for compliance (HIPAA, GDPR, PCI-DSS, etc.).
• ✅ Zero-Trust Logic: Strict separation of facts ([FACT]) and inferences ([INFERENCE]) to prevent hallucinations.
• ✅ Interactive Self-Evaluation: Built-in 1-10 scoring check with self-correction logic.
• ✅ 15 worked few-shot scenarios: Covering standard runs, edge cases, out-of-scope requests, and validation failures.
• ✅ 55-case appendix: Context-rich manual pages to help models maintain their persona during long execution sessions.

No setup. No API keys. No dependencies. Just find what you need and use it.

---

📊 What's Inside

Dimension	Count	Details
Departments	12	Sales, Finance, HR, Marketing, Engineering, Customer Support, Operations, Product, Legal, Security, IT Admin, Design
Roles per Dept	10	120 unique professional agent profiles
Industry Verticals	12	SaaS, FinTech, E-commerce, HealthTech, Real Estate, EdTech, Logistics, PropTech, AgTech, BioTech, CyberSecurity, InsurTech
Framework-Agnostic Prompts	1,440	raw_prompt.md files (2,100+ lines each) for any LLM or client
Framework & Model Configs	17,280	Optimized files for CrewAI, LangGraph, AutoGen, and Vercel AI SDK
Total Configurations	18,720	Every combination of the above

---

📂 How Files Are Organized

skins/
├── sales/
│   ├── saas/
│   │   └── lead_qualifier/
│   │       ├── crewai_openai_optimized.py
│   │       ├── crewai_anthropic_optimized.py
│   │       ├── langgraph_llama_optimized.py
│   │       ├── autogen_openai_optimized.py
│   │       ├── vercel_ai_sdk_anthropic_optimized.ts
│   │       └── ... (12 files per role)
│   ├── fintech/
│   ├── healthtech/
│   └── ... (12 industries)
├── finance/
├── hr/
├── marketing/
├── engineering/
├── customer_support/
├── operations/
├── product/
├── legal/
├── security/
├── it_admin/
└── design/

Path formula: skins/{department}/{industry}/{role}/{framework}_{model}_optimized.{py|ts}

---

⚡ Quick Start

Option 1: Browse and Copy

Navigate the skins/ folder, find the file matching your use case, and copy the contents into your project.

Option 2: Use the Interactive CLI

We ship a zero-dependency search tool at the root of this repo:

List all departments:

python skins.py list departments

List all industries:

python skins.py list industries

Search by keyword:

python skins.py search expense
python skins.py search vulnerability
python skins.py search onboard

Fetch a specific skin:

python skins.py get sales saas lead_qualifier crewai openai
python skins.py get security healthtech vulnerability_auditor langgraph anthropic
python skins.py get hr edtech candidate_screener autogen llama

---

🛠️ Example Output

Running python skins.py get finance fintech expense_reconciler crewai anthropic gives you:

from crewai import Agent, Task, Crew

def initialize_agent(tools):
    return Agent(
        role="Expense Reconciler",
        goal="Validate employee expense submissions against company travel and spend policies.",
        backstory="A seasoned Expense Reconciler with deep expertise in the FinTech vertical.",
        tools=tools,
        verbose=True,
        memory=True
    )

# Full system prompt with PCI-DSS compliance guardrails,
# FinTech domain vocabulary, and Anthropic XML-tag formatting
# is included in the file footer.

---

🏗️ Supported Frameworks

Framework	Language	Use Case
CrewAI	Python	Multi-agent teams with role-based collaboration
LangGraph	Python	Stateful graph-based agent orchestration
AutoGen	Python	Conversational multi-agent systems
Vercel AI SDK	TypeScript	Edge-deployed AI applications

---

🔒 Built-In Compliance Standards

Every skin includes industry-appropriate guardrails:

Industry	Standard	What the Guardrail Does
SaaS	GDPR & SOC2	Blocks PII exposure, enforces data minimization
FinTech	PCI-DSS & SEC	Blocks raw credit card numbers, enforces transaction audits
HealthTech	HIPAA & HITECH	Blocks PHI in logs, enforces encryption rules
EdTech	FERPA & COPPA	Blocks minor profiles without parent auth
E-commerce	PCI-DSS	Verifies price formats, blocks negative stock
BioTech	FDA Part 11	Blocks modification of locked trial records
CyberSecurity	ISO 27001 & NIST	Enforces log integrity, blocks write to event archives
InsurTech	NAIC Rules	Audits claim calculations against policies
Real Estate	Fair Housing Act	Enforces zero-bias descriptions
Logistics	FMCSA	Enforces weight limits and transport regulations
PropTech	Tenant Rights	Blocks unsigned lease modifications
AgTech	USDA & EPA	Enforces chemical and pesticide tracking

---

🤝 Contributing

Want to add a new role, industry, or framework? See CONTRIBUTING.md.

---

📄 License

MIT — use it however you want. See LICENSE.