Agent Skills

Overview

Agent Skills are small, self-contained instruction packs that teach Claude (and other skill-aware agents) how to perform a specific task. Webhook Relay publishes an open-source set of Skills that show agents how to drive the platform with the relay CLI and API — forwarding webhooks, transforming them, debugging incoming requests, exposing services over tunnels, scheduling recurring webhooks, and receiving, parsing and transforming inbound email.

The Skills are maintained on GitHub:

• github.com/webhookrelay/skills

Use that repository to browse the source, open issues, or contribute new skills.

Zero-signup first step: capture a webhook

Before installing anything or creating an account, get a public webhook URL from the free Webhook Bin API — one call, no account, no key:

B=https://bin.webhookrelay.com
BIN=$(curl -s -X POST $B/v1/bins | jq -r .id)   # create a bin — no account, no key
echo "$B/v1/webhooks/$BIN"                       # give this URL to any webhook sender
curl -s "$B/v1/bins/$BIN" | jq '.requests'       # read back exactly what arrived

The bin records every request — method, headers, query string and body — and can mock responses and verify HMAC signatures. Bins are public and expire after about 48 hours, so never point one at secrets. This is the API the webhook-debug skill wraps, and the first move in agentic-webhook-testing.

When you need real delivery — forwarding those webhooks to localhost, a private network or Jenkins, with retries and transforms — create a free account (no credit card) and continue with the Quick start.

Quick start

Install every skill with the skills.sh CLI:

npx skills add webhookrelay/skills

That drops the skills into your agent's skills directory. From then on you can ask your agent in plain language — it picks the right skill and runs the relay commands for you:

See all the ways to install — single skill, Claude Code plugin, ClawHub, or manual — below.

Skills vs. MCP

Skills and the MCP server solve related but different problems, and they work well together:

• Skills are instructions. They teach an agent how to use the relay CLI/API correctly — the right commands, payload shapes, and workflows. They run wherever your agent has a shell.
• MCP is a live connection. It gives an agent typed tools to manage buckets, inspect webhook logs, and work with transform functions directly against your account.

Install the Skills for guided, command-line workflows; connect MCP when you want the agent to call Webhook Relay tools directly.

Available skills

Outbound webhook playbooks

To send webhooks to your own customers, install the outbound-webhooks skill above. To move them to the Outbound Gateway from another provider, also point your agent at the hand-written migration playbook for that provider. Each one lists the exact API calls, the checks to run and what the agent must never do: it imports your customers' existing signing secrets where their format allows (Svix secrets always do), adds dual publishing, and stops for your approval before any customer traffic moves:

The playbook index helps an agent pick the right one, and outbound-webhooks.md is the agent-facing setup reference. Connect the MCP server so the agent can use the outbound tools directly. The Outbound Gateway is available on every plan, Free included.

Installation

The fastest way to install is with the skills.sh CLI:

npx skills add webhookrelay/skills

Add -g to install globally for all your agents:

npx skills add webhookrelay/skills -g

Install a single skill

Pass --skill <name> to install just one:

npx skills add webhookrelay/skills --skill webhook-debug

Claude Code plugin marketplace

The repository doubles as a Claude Code plugin marketplace. Add it, then install the plugin:

/plugin marketplace add webhookrelay/skills
/plugin install webhookrelay-skills

ClawHub

The skills are also published on ClawHub, the open registry for AI agent skills. Install the bundled Webhook Relay skill with the openclaw CLI:

openclaw skills install webhook-relay

See Webhook Relay on ClawHub for what the registry does and how the skill fits.

Manual install

You can also copy any skills/<name>/ folder from the repository into your agent's skills directory.

Prerequisites

The webhook-debug skill works against the public bin API with nothing but curl — no account or key needed. agentic-webhook-testing prefers a connected Webhook Relay MCP server so the agent can use the typed send, wait, delivery-log and function-log tools.

For CLI fallback and skills that operate a local runtime, install and authenticate the relay CLI:

1. Install it — see CLI installation.
2. Log in: relay login (or set the RELAY_KEY / RELAY_SECRET environment variables).
3. Verify: relay bucket ls.

Concepts at a glance

• Bucket — groups inputs and outputs.
• Input — a public HTTPS endpoint that receives webhooks.
• Output — a destination requests are relayed to; internal (delivered by a running agent, e.g. localhost) or public (delivered server-side).
• Function — server-side JavaScript/Lua that transforms requests/responses.
• Tunnel — a public hostname that proxies inbound HTTP/TCP to a local or internal service.
• Cron — a scheduled, recurring webhook.

See Getting Started for the full model.

Issues and contributions

The Skills are open source. To report a problem, request a new skill, or read the code, use the GitHub repository:

• github.com/webhookrelay/skills — browse the code and open an issue.

Links

• Webhook Relay Skills on GitHub
• skills.sh CLI — what it is and how to install with it
• Webhook Relay on ClawHub
• MCP server
• Functions reference
• CLI installation
• Dashboard
• LLM-friendly docs index — append .md to any docs URL for plain markdown, or see llms.txt