Amazon SES Agent Skills: AI-Assisted Email Development Explained

Ask an AI coding agent to “send my first email with Amazon SES” and it will often hand you code that fails. The agent reaches for the deprecated V1 API, skips identity verification, and forgets that a brand-new account is locked in a sandbox that cannot email real recipients. You paste the code, run it, and spend the next hour reading AWS documentation to find out why nothing arrived. The model was confident. It was also wrong.

On June 4, 2026, the Amazon SES team shipped a fix for exactly this problem: Amazon SES Agent Skills, an open-source set of skills that feed AI coding agents the correct context before they write a single line. The skills work with Claude Code, Kiro, and any tool that supports the open Agent Skills format. This guide explains what they do, how to install them, and where they fit if you are choosing a transactional email provider in the first place.

What an agent skill actually is

An agent skill is a structured context package that teaches an AI agent how to use one specific service correctly. Instead of relying on training data — which goes stale the moment an API changes — the agent loads validated patterns, a list of common mistakes, and step-by-step workflows scoped to the task in front of it.

The Amazon SES release ships two skills, because SES is really two products with two different API clients:

The aws-ses skill covers sending email — transactional messages, marketing campaigns, and notifications — using the sesv2 client. The aws-mail-manager skill covers receiving and processing inbound mail (routing, filtering, archiving, SMTP relay) using the mailmanager client. These are separate APIs, and one of the most frequent agent errors is mixing them up or falling back to the legacy ses client for sending.

Each skill is built from three parts. A SKILL.md file describes what the skill does, when to use it, and which mistakes to avoid. A references/ folder holds task-specific guides for things like identity verification, configuration sets, and troubleshooting. A scripts/ or examples/ folder contains working code in Python, Node.js, and Java. The agent pulls in only the slice it needs — ask about sending and it loads the sending guides, ask about archiving inbound mail and it loads the archive reference.

Prerequisites

Before the skills are useful, you need a few things in place:

1. An AI coding agent that supports the Agent Skills format. Claude Code and Kiro both qualify, as does any agent compatible with the open format.
2. An AWS account with Amazon SES access. The skills assume SES is enabled in your chosen region.
3. AWS credentials configured through environment variables, a shared credentials file, or an IAM role. The agent generates code; it does not manage your authentication for you.
4. The SDK for your language — boto3 for Python, @aws-sdk/client-sesv2 for Node.js, or software.amazon.awssdk:sesv2 for Java.

How to install the skills

Installation runs through the skills CLI. To add both skills at once:

npx skills add amazon-ses/skills

To install only the one you need:

npx skills add amazon-ses/skills --skill aws-ses
npx skills add amazon-ses/skills --skill aws-mail-manager

Once installed, the skill activates on its own. When you ask your agent an email-related question, it recognizes the context and loads the relevant guidance — no flag or manual trigger required.

What the workflow looks like in practice

Here is the difference the aws-ses skill makes. Ask your agent, “Help me send my first email with Amazon SES.”

Without the skill, you might get code built on the V1 API, with no identity verification and no configuration set. With the skill loaded, the agent follows the order of operations SES actually requires:

1. Confirms your sending identity is verified (a domain or an email address).
2. Checks whether the account is still in the sandbox and recommends simulator addresses for safe testing.
3. Creates a configuration set so you get event tracking from the start.
4. Sets up a tenant to isolate the workload.
5. Generates V2 API code with real error handling.

The Python it produces looks like this:

import boto3
from botocore.exceptions import ClientError

client = boto3.client('sesv2', region_name='us-east-1')

try:
    response = client.send_email(
        FromEmailAddress='sender@example.com',
        Destination={'ToAddresses': ['success@simulator.amazonses.com']},
        Content={
            'Simple': {
                'Subject': {'Data': 'Hello from Amazon SES'},
                'Body': {'Text': {'Data': 'This email was sent using Amazon SES V2 API.'}}
            }
        },
        ConfigurationSetName='my-config-set',
        TenantName='my-tenant'
    )
    print(f"Message sent: {response['MessageId']}")
except ClientError as e:
    print(f"Send failed: {e.response['Error']['Code']} - {e.response['Error']['Message']}")

Three details matter here. The client is sesv2, not the legacy ses. The send includes a configuration set for observability. The destination is a simulator address (success@simulator.amazonses.com), so you can test delivery, bounces, and complaints without touching a real inbox or denting your sender reputation.

The Mail Manager side

The aws-mail-manager skill handles inbound mail, and it enforces an architecture that is easy to get wrong by hand. The pipeline runs: Internet → Ingress Point → Traffic Policy → Rule Set → Action.

The dependency order is the trap. Resources have to be created back to front — the traffic policy and rule set must exist before the ingress point that references them. The skill also teaches the correct condition syntax (union types with exactly one key per object) and makes the agent wait for the ingress point to reach ACTIVE status before recommending any DNS changes. Get that sequence wrong manually and you end up with an ingress point pointing at resources that do not exist yet, which fails in ways the error messages do not explain well.

Common mistakes the skills prevent

The value of these skills is mostly defensive. They stop the errors that cost the most debugging time:

• Using the V1 API. The legacy ses client still exists and still appears in old tutorials, so models reproduce it constantly. The skill steers every send to sesv2.
• Skipping identity verification. SES will not send from an unverified domain or address. Agents that skip this step generate code that throws on the first call.
• Ignoring the sandbox. Code that works against simulator addresses can still fail against real recipients while the account is sandboxed. The skill flags this before you waste time hunting for a bug that is really an account-status issue.
• No configuration set. Without one, you get no event tracking — no opens, bounces, or complaints — and retrofitting observability later is more work than adding it up front.
• Mixing send and receive clients. Reaching for mailmanager to send, or sesv2 to process inbound mail, produces code that does not do what you asked.

Where Amazon SES fits — and where it doesn’t

The skills make SES easier to build against, but they do not change what SES is: raw sending infrastructure with no marketing interface. Before you commit, it helps to know the tradeoff.

SES pricing is the draw. The base rate is $0.10 per 1,000 emails, charged per recipient, with no monthly minimum. AWS includes a free tier of up to 3,000 message charges per month for the first 12 months. Dedicated IPs run $24.95/month each on the standard option, and the Virtual Deliverability Manager adds $0.07 per 1,000 emails. At one million emails a month, base sending costs around $100 — far below what most managed platforms charge at that volume. Figures are from Amazon’s SES pricing page; confirm current rates before budgeting.

The weaknesses you need to know. SES has no campaign builder, no template gallery, and no subscriber management — it is an API and nothing more. You own domain verification, DKIM, DMARC, bounce handling, and suppression lists. Support is not free either: meaningful help starts at $29/month on the AWS Developer plan, and the free tier excludes delivery troubleshooting. The agent skills shrink the setup learning curve, but they do not give SES a dashboard your marketing team can use.

For teams that want deliverability without the operational load, a managed transactional provider is often the better call. Postmark, for example, runs separate IP pools for transactional and bulk mail and publishes its deliverability numbers openly.

Postmark’s free tier is 100 emails/month with no expiry, and paid plans start at $15/month for 10,000 emails. Its weakness is the flip side of its discipline: it bans cold outreach and promotional bulk sending, and at multi-million-email volumes its per-message overage rates lose to SES on pure cost. Pricing is from our last verified check on May 5, 2026 — see the Postmark pricing page for current numbers.

If you want to weigh the cost curve against the engineering effort, our Amazon SES review and SES pricing breakdown lay out the full picture, and the best transactional email services roundup ranks SES against the managed alternatives.

What success looks like

You know the skills are working when your agent stops producing code you have to rewrite. A first send should use sesv2, target a simulator address, attach a configuration set, and handle the ClientError path. Your account should already have a verified identity and a pending or approved production-access request. Event data — bounces, complaints, deliveries — should be flowing into wherever you pointed your configuration set, because the tracking was wired in from the first commit rather than bolted on after the first deliverability scare.

That is the real win. The skills do not make SES simple; they make the AI assistant building on top of it accurate. For a developer choosing between rolling their own SES integration and paying for a managed service, that accuracy narrows the gap — though, as always with SES, authentication and deliverability remain your job. Whatever provider you land on, set up SPF, DKIM, and DMARC before your first production send; our email deliverability guide walks through it.

The skills are available now on GitHub under the Apache-2.0 license, and the open Agent Skills format means other providers can ship their own. If transactional email is part of your stack, this is a low-cost upgrade to how you build it.

Sources

1. AWS — Getting started with Amazon SES Agent Skills — accessed 2026-06-11
2. amazon-ses/skills — GitHub repository — accessed 2026-06-11
3. Amazon SES — Pricing — accessed 2026-06-11
4. Postmark — Pricing — accessed 2026-06-11
5. Agent Skills format — accessed 2026-06-11