Skip to main content

​
What Are Skills?

Skills are custom HTTP-based tools that extend what your agents can do. While RelayHub includes built-in tools for file analysis, data visualization, and web search, skills let you connect agents to any external API or service your business uses. A skill might call your CRM to look up a customer, query your internal reporting API for sales data, post a message to Slack, or trigger a deployment pipeline. Once created, a skill can be enabled on any agent, giving that agent the ability to take real actions in your systems.

​
The Skill Hub Interface

Open Skill Hub from the sidebar to browse all available skills. Each skill card shows:
  • Name and description — what the skill does
  • HTTP method and endpoint — the API it calls
  • Required credentials — whether the skill needs authentication
  • Enabled status — whether the skill is currently active

Browse Skills

View all skills available to your organization, filter by category, and see which agents use each skill.

Create Skills

Build new skills with the guided Skill Builder wizard or configure them manually.

​
Building a Skill with the Skill Builder

The Skill Builder is an AI-guided wizard that walks you through creating a skill step by step.
1

Describe the integration

Tell the Skill Builder what you want the skill to do in plain language. For example: “Look up a customer by email in our HubSpot CRM and return their company name, deal stage, and last activity date.” The builder generates a draft configuration based on your description.
2

Configure the HTTP request

Review and adjust the generated endpoint URL, HTTP method (GET, POST, PUT, DELETE), headers, query parameters, and request body. The builder pre-fills these based on your description, but you can modify any field.
3

Define parameters

Specify what inputs the skill accepts. These become the parameters the LLM fills in when it decides to use the skill. Each parameter has a name, type, description, and required flag. Clear descriptions help the model supply correct values.
4

Add credentials

If the API requires authentication, configure credentials. RelayHub supports API keys (header or query parameter), Bearer tokens, and basic auth. Credentials are encrypted and stored securely — they are never exposed to the LLM or included in conversation history.
5

Test the skill

Run a test call with sample parameters to verify the skill works. The builder shows the full request and response so you can debug any issues before saving.
6

Save and enable

Save the skill and enable it on one or more agents. The skill is immediately available the next time someone chats with those agents.

​
Credentials Management

Skills that call authenticated APIs need stored credentials. RelayHub manages these securely.
Credentials are encrypted at rest and never included in LLM prompts or conversation logs. The model only sees the skill name, description, and parameters — never the authentication details.
You can manage credentials from the Skill Hub:
  • Add credentials — store API keys, OAuth tokens, or username/password pairs
  • Rotate credentials — update a key without editing every skill that uses it
  • Revoke credentials — immediately disable access for a credential set
Each credential set can be shared across multiple skills. For example, if three skills call the same API with the same key, they reference one credential entry. Rotating that key updates all three skills at once.

​
Testing Skills

Before enabling a skill on a production agent, test it thoroughly:
  1. Use the built-in test panel in the Skill Builder to send requests with sample data
  2. Check the response to make sure the skill returns the data format the LLM can interpret
  3. Verify error handling by testing with invalid inputs — the skill should return clear error messages, not raw stack traces
Write the skill description from the LLM’s perspective. Instead of “Calls the HubSpot API”, write “Look up a customer record by email address and return their company, deal stage, and last activity.” The description is what the model reads when deciding whether to use the skill.

​
Enabling Skills on Agents

Once a skill is saved, you can enable it on any agent:
  1. Open the agent in the Agent Hub editor
  2. Scroll to the Tools section
  3. Toggle on the skill under Custom Skills
  4. Save the agent
The agent can now invoke that skill during conversations when the user’s request matches the skill’s purpose. The model reads the skill name, description, and parameter definitions to decide when and how to call it.
Skills are available to all agents in your organization. Scope is controlled at the agent level — you choose which agents get access to which skills. A sensitive skill (like one that modifies production data) should only be enabled on agents used by authorized team members.

​
Example Use Cases

SkillWhat It Does
CRM LookupQueries your CRM by customer name or email, returns deal info
Ticket CreatorOpens a support ticket in your helpdesk system
Inventory CheckQueries product availability from your ERP
Slack NotifierPosts a formatted message to a Slack channel
Deploy TriggerKicks off a CI/CD pipeline for a given branch