agent-email-inbox Skill (Resend) for Claude: What It Does and How to Install It

Makes Claude build an email inbox for an AI agent that checks who is writing before it acts on a message.

Installs: 13.8K; on skills.sh
Stars: 208; on GitHub
Files: 3; besides SKILL.md
Updated: Oct 2026; last repo commit

This skill from Resend walks Claude through an inbox for an AI agent on Resend: pick a security level, verify a domain, create the inbox, add a webhook endpoint and reply in the thread. It sets five security levels, from a strict sender allowlist to human approval.

It loads for any system where email content triggers actions, even without an agent. It needs a Resend account, a RESEND_API_KEY with full access, and a public webhook URL for Resend to call.

Good for

Developers building an AI agent that reads and answers email
Teams turning inbound email into tasks or support replies
Anyone who must keep untrusted email from steering an agent
Engineers wiring Resend webhooks to a server

Install agent-email-inbox

Claude Code

Copy this prompt.

Install the agent-email-inbox skill in Claude Code. Run this command:

/plugin marketplace add resend/resend-skills
/plugin install resend@resend-skills

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

Pick one route. The plugin route installs the resend plugin: 5 skills (counted from the repo tree) and a hosted Resend MCP server entry from mcp.json, which the README says signs in with OAuth. The skills routes install just this skill. SKILL.md suggests the resend skill for full sending docs, but does not require it.

Built from the package data we hold. Not yet tested by us. Source.

What's in the skill

When it loads

Use when building any system where email content triggers actions — AI agent inboxes, automated support handlers, email-to-task pipelines, or any workflow processing untrusted inbound email. Always use this skill when the user wants to receive emails and act on them programmatically, even if they don't mention "agent" — the skill contains critical security patterns (sender allowlists, content filtering, sandboxed processing) that prevent untrusted email from controlling your system.

Sections in SKILL.md

Overview
Why Webhook-Based Receiving?
Architecture
SDK Version Requirements
Quick Start
Before You Start: Account & API Key Setup
First Question: New or Existing Resend Account?
Creating API Keys Securely
API Key Permissions
Domain Setup
Option 1: Receiving on the Domain (Recommended)
Option 2: Forwarding
Create the Inbox
Security Levels
Level 1: Strict Allowlist (Recommended)
Security Best Practices

Key facts

Author: Resend (vendor's own repo; named in SKILL.md metadata)
License: MIT
Installs on skills.sh: 13,759 (Nov 2, 2026)
GitHub stars: 208 for the whole resend/resend-skills repo (Nov 2, 2026)
Last repo commit: Oct 28, 2026
What it includes: 4 files including SKILL.md: 3 reference guides. No scripts
skills.sh audits: Gen Agent Trust Hub: pass; Socket: pass; Snyk: warn (Nov 2, 2026)
Needs: A Resend account, a custom domain, and a full access RESEND_API_KEY. Inboxes need the Node.js resend package 6.34.1 or later, or Python resend 2.50.0 or later
Works in: Claude Code and Cursor as a plugin, and any Agent Plugins client, per the README
Part of: The resend plugin (version 1.0.5) in resend/resend-skills: 5 skills counted from the repo tree, plus a hosted Resend MCP server entry

What it includes

Eight steps from choosing a security level to connecting the agent
Write · Quick start
Dedicated account advice, and why a sending access key is not enough
SKILL.md says not to paste keys in chat.
Account and API key setup
Verify DKIM and SPF, enable receiving and add an MX record, or use forwarding
Write · Domain setup
resend.inboxes.create with an address, name and fromName
Write · Create the inbox
Only listed senders are processed; recommended for most uses
Level 1: Strict allowlist
Anyone at trusted domains can interact
Level 2: Domain allowlist
Accept from anyone and filter unsafe patterns
Level 3: Content filtering
Process all email with restricted agent capabilities
Level 4: Sandboxed processing
A person approves untrusted actions
Level 5: Human in the loop
Next.js and Express examples that verify the signature on the raw body
Network · Webhook endpoint
resend.inboxes.threads.emails.reply sets recipients and threading headers
Write · Reply in the thread
Use email.received and emails.receiving.get when there is no custom domain
Network · Receiving without an inbox
Register the webhook by API; tunnel options include Tailscale Funnel and ngrok
Includes curl | sh and sudo commands for Tailscale.
Runs code · Webhook setup and tunnels
Per-sender rate limits, content length limits, stripping quoted threads, troubleshooting
Advanced patterns

From the official docs

Permissions

High risk

What it can do

Read your data
Create and change things
Delete things
Read-only mode

Watch out for

Untrusted email can steer an agent
SKILL.md warns that anyone can email instructions to an unprotected agent. Choose a security level before any email is processed, and check Reply-To against the allowlist before sensitive replies. github.com
It needs a full access API key
SKILL.md says a sending access key returns 401 for reading inbox email and registering webhooks. It advises a dedicated Resend account so a leaked key cannot reach other projects. github.com
The tunnel guide runs a downloaded installer
webhook-setup.md installs Tailscale with curl -fsSL https://tailscale.com/install.sh | sh and then runs sudo tailscale up and sudo tailscale funnel 3000. That makes your local port public, so read the script first. github.com

Security advisories

No published security advisories as of Nov 2, 2026.

Technical details

Turning on read-only	No read-only mode. Use the strictest security level and a dedicated Resend account.
Talks to	The Resend API (SDK calls). tailscale.com install script, if you follow the Tailscale steps. Your own public webhook URL, called by Resend
Sign-in and keys	SKILL.md says to keep the key on the server only, never pass it to the AI model, and not to paste it in chat; it suggests a .env file or a secrets manager.
Scopes	Code Claude writes with a full access RESEND_API_KEY can create inboxes and webhooks and reply to email The webhook guide exposes a local port to the internet through a tunnel webhook-setup.md lists install commands for Tailscale, ngrok and cloudflared, and a vercel deploy command

Questions

How do I install agent-email-inbox?

Use the install box at the top of this page. The plugin route installs all 5 skills in the Resend repo plus a hosted MCP server entry, and the skills CLI route installs just this skill.

What does the agent-email-inbox skill do?

It helps Claude set up an email inbox for an AI agent with security built in. Claude follows its steps to pick a security level, create the inbox and webhook, and reply in the thread.

Is this Resend's official skill?

Yes, it is in the vendor's own repo, resend/resend-skills. SKILL.md names resend as author, and the README marks it as "Authored here".

Why does an agent inbox need security levels?

SKILL.md says an agent's inbox receives untrusted input, and anyone can email instructions that an unprotected agent will execute. It offers five levels, with a strict sender allowlist recommended for most uses.

Do I need a custom domain and a paid plan?

The inbox flow uses a custom domain you have added to Resend, or forwarding. SKILL.md says a 403 when you create an inbox means your plan has no inbox left, and it describes receiving without an inbox as the fallback.

Which API key does it need?

A full access key. SKILL.md says a sending access key can only reply, forward and manage drafts, and gets a 401 when reading inbox email or registering webhooks. It also advises a dedicated Resend account for the agent.

agent-email-inbox or the resend skill?

Use this one when email content triggers actions. SKILL.md suggests the resend skill for full sending docs, and the resend skill points back here for untrusted input. They are separate installs.

What values does SECURITY_LEVEL take?

The frontmatter lists strict, moderate and permissive. The environment variable example in SKILL.md lists strict, domain, filtered and sandboxed. The two lists differ, so set the level in your own code and check it against the five levels in the guide.

Is it safe?

The folder bundles no scripts, but the setup touches real systems. It uses a full access Resend key, a public webhook URL and, for local work, tunnel tools that the reference guide installs with sudo. Follow its allowlist advice before the agent acts on any message.

Information

updated Nov 2, 2026

Repository: resend/resend-skills
Folder: skills/agent-email-inbox

resend/resend-skills skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 5

agent-email-inbox SKILL.md · github.com
resend-skills README · github.com
security-levels.md · github.com
webhook-setup.md · github.com
advanced-patterns.md · github.com