Courier Skills

Courier Skills is a public repository of structured guidance that AI coding agents use when building notifications. Install it once and your agent gets channel-specific rules, compliance constraints, reliability patterns, and code examples for the full Courier platform. Works with Cursor, Claude Code, Windsurf, Cline, and any AI tool that supports agent skills.

Installation

Cursor (global)
Cursor (project)
Claude Code
Other

Available in all projects:

git clone https://github.com/trycourier/courier-skills.git ~/.cursor/skills/courier-skills

Available in one project only:

git clone https://github.com/trycourier/courier-skills.git .cursor/skills/courier-skills

git clone https://github.com/trycourier/courier-skills.git ~/.claude/skills/courier-skills

Claude Code discovers skills from ~/.claude/skills/ automatically via SKILL.md frontmatter.

Clone the repo into whatever skills directory your AI assistant supports, or point it at the SKILL.md file manually.

git clone https://github.com/trycourier/courier-skills.git

What Your Agent Learns

Category	Coverage
7 Channels	Email, SMS, push, in-app inbox, Slack, Microsoft Teams, WhatsApp; each with deliverability rules, compliance requirements, and code examples
5 Transactional Types	Authentication (OTP, password reset), orders, billing, appointments, account notifications; timing requirements, security rules, and templates
6 Growth Types	Onboarding, adoption, engagement, re-engagement, referral, campaigns; consent requirements, frequency limits, and exit conditions
10 Cross-Cutting Guides	Multi-channel routing, preferences, compliance (GDPR/TCPA/CAN-SPAM), reliability, batching, throttling, CLI usage, and reusable patterns

How It Works

The skill is organized around a routing file (SKILL.md) that agents read first. Based on the task at hand, it directs the agent to read only the 1-2 files relevant to that task. Each resource file follows a consistent structure:

Quick Reference at the top with hard rules, common mistakes, and copy-paste templates
Detailed guidance with explanations, examples, and edge cases
Related links to other relevant files

This means agents read the minimum necessary context rather than loading the entire skill.

Example: agent building an OTP flow

Agent reads SKILL.md, finds the routing table
Routing table says: OTP/2FA -> read authentication.md + sms.md
Agent reads those two files, gets: 6-digit codes, 5-10 min expiry, SMS primary with email fallback, TCPA rules, idempotency key pattern, rate limits
Agent generates code that follows all constraints

Code Examples

The skill includes TypeScript, Python, CLI, and curl examples for key patterns:

# Send with idempotency (CLI)
courier send message \
  --message.to.user_id "user-123" \
  --message.template "ORDER_CONFIRMATION" \
  --message.data '{"orderId": "12345"}'

// Send with idempotency (TypeScript)
await courier.send({
  message: {
    to: { user_id: "user-123" },
    template: "ORDER_CONFIRMATION",
    data: { orderId: "12345" }
  }
}, {
  idempotencyKey: "order-confirmation-12345"
});

# Send with idempotency (Python)
client.send(
    message={
        "to": {"user_id": "user-123"},
        "template": "ORDER_CONFIRMATION",
        "data": {"orderId": "12345"},
    },
    idempotency_key="order-confirmation-12345",
)

Universal Rules

The skill enforces these constraints across all generated code:

Never send promotional content in transactional notifications
Never batch or delay OTP, password reset, or security alerts
Never send SMS without TCPA-compliant consent records
Always use idempotency keys for transactional sends
Always respect quiet hours (10pm-8am local) unless critical
Always use method: "single" unless the notification warrants all channels

Platform

Developer Tools

Help

Installation

What Your Agent Learns

How It Works

Example: agent building an OTP flow

Code Examples

Universal Rules

What’s Next

Build with AI

GitHub

​Installation

​What Your Agent Learns

​How It Works

​Example: agent building an OTP flow

​Code Examples

​Universal Rules

​What’s Next

Build with AI

GitHub

Installation

What Your Agent Learns

How It Works

Example: agent building an OTP flow

Code Examples

Universal Rules

What’s Next