A Skill is a playbook that tells the agent when to activate and how to carry out one feature, from the trigger through every step, including any API call to make along the way. You describe the feature in plain language, and Chatropic generates the playbook for you to review and refine. Skills sit one level above actions. An action is a single tool, such as one API call. A skill is the full behavior around it: the conditions to check, the questions to ask, the API to call, the tags to apply, and when to escalate. When you save a skill that contains a webhook step, Chatropic creates (or updates) the linked HTTP action under Actions > Available actions automatically.

​
Create a skill

1

Open Skills

In the sidebar, go to Agent settings > Skills and click New skill.
2

Pick a trigger

In the When selector, choose what starts the skill: Customer sends a message or Customer is on a specific page.
3

Describe the skill

Write a prompt that covers the trigger, any conditions, the API call if one is needed, and the reply you want. For example:
When the customer is on the checkout page and asks for a discount, offer them a 5% discount code “lastminute5”, and tag the chat as “promotions”.
Open What can I use? below the prompt box for a reminder of what the generator understands: triggers, conditions, API lookups, tags, and replies. Mention URLs and authentication if the skill should call an external API.
4

Generate

Click Create skill. Chatropic drafts the playbook and opens it in the editor. New skills start as Draft, so nothing goes live until you activate it.
Not sure where to start? The New skill page also offers ready-made templates in categories such as External Connections, Ecommerce, Customer Support, and Operations. Clicking a template clones it into your workspace as a draft you can edit.

​
Review and edit the playbook

The editor shows the generated skill as an ordered list of steps under a When to use this procedure trigger section. Check three things before activating:
  1. When to use matches the situations you intended, and nothing broader.
  2. Steps read like instructions you would give a new support agent.
  3. Webhook details (URL, headers, body) point at the right endpoint.

​
Step types

Build or adjust the flow with the step picker: Steps can also attach behaviors: use an action from your catalog, tag the conversation (tagged messages can trigger handoff rules), or read an attribute from the conversation context.

​
Webhook steps and linked actions

A webhook step is how a skill calls your API. Configure it on the step panel:
  • URL with {{variable}} placeholders for dynamic values, for example https://api.example.com/records/{{record_id}}
  • Method and headers, including saved credentials as {{token.<name>}} from Manage tokens
  • An optional JSON body template, also using {{variable}} placeholders
  • An optional output widget that renders the response as a card or list in chat
Use the built-in Test button to probe the endpoint with sample values before going live. If the response does not match the attached widget’s shape, the panel warns you.
When you save the skill, each webhook step creates or updates a linked HTTP action under Actions > Available actions. Edit the webhook from the skill rather than editing the linked action directly, so the two stay in sync.

​
Activate the skill

The editor has two save buttons: The status badge on the Skills list shows which skills are live. Deactivated skills stay in the list as drafts, so you can iterate without deleting anything.
The number of skills per workspace depends on your plan. The dashboard tells you when you reach the limit.

​
Test with simulations

Each skill has a simulation panel where you define test cases: a scenario the customer might raise, and what the skill should do. Run cases one at a time or click Run all, and each case records a pass or fail with a timeline of what the agent did. If your webhook needs authentication, add simulation credentials so test runs can call the API. Simulations run against the skill definition, so use them the way you use unit tests: after every meaningful edit and before activating.

​
What success looks like

  • The skill shows Live on Agent settings > Skills
  • Its linked action appears under Actions > Available actions
  • In Playground, a message matching the trigger runs the playbook and, for webhook skills, calls your API
  • Simulation cases pass

​
Next steps

Worked example

Three skills powering a grocery shopping agent, end to end.

Manage tokens

Credentials for webhook steps.

Actions overview

How linked actions fit into the catalog.

Escalate to human

Where Escalate steps hand off, plus keyword and frustration triggers.