Skip to main content
Skills are folders of instructions and reference files that teach your agent how to handle specific tasks — consistently, without repeating yourself in every system prompt. Attach a skill to an agent node and it instantly gains that expertise.

​
Prerequisites

​
Part 1: Use a Skill from the Catalog

The fastest way to get started is to fork a published skill. For example, Telegram Bot Creator automates Telegram bot creation and deployment to Digital Ocean or Hetzner Cloud using python-telegram-bot + Docker Compose — supports workflow bots, inline bots with Telegram Stars payments, and Mini App bots.

​
Step 1: Find and Fork the Skill

  1. Go to app.splox.io and click Skills in the left sidebar
  2. Switch to the Catalog tab and search for telegram-bot-creator
  3. Click Fork — Splox creates an independent copy in your library
Skills catalog with telegram-bot-creator

​
Step 2: Attach to Your Agent

  1. Open your workflow and click the Agent node
  2. In the config panel under the Model tab, find the Skills field
  3. Select your forked skill from the dropdown
Attaching a skill to an agent node

​
Step 3: Run It

Open the Chat panel and ask your agent to create a Telegram bot. The agent reads the skill’s instructions on demand and follows the deployment steps automatically.
See a real example — the agent deploys a full Telegram bot with Telegram Stars payments to a Hetzner server in one conversation.
You can fork any published skill and customize the files freely. Your copy is fully independent.

​
Part 2: Create Your Own Skill

​
Skill Structure

A skill is a folder of files following a simple convention: 
my-skill/
├── SKILL.md          # Main instructions (required)
├── references/       # Deep reference docs (loaded on demand)
└── scripts/          # Executable code (optional)
The agent loads content in layers: it always reads the YAML frontmatter, loads SKILL.md when the skill is relevant, and navigates into references/ only when it needs deeper context. This keeps token usage lean.

​
SKILL.md Format

The frontmatter is the most important part — it’s how the agent decides when to use the skill: 
---
name: brand-voice
description: Writes content in our company tone of voice. Use when user asks
  to write announcements, emails, blog posts, or any public-facing copy.
---

## Instructions

1. Always use an active, direct voice
2. Keep sentences under 20 words
3. Avoid jargon — write for a general audience
...
The description field drives triggering. Include what the skill does and when to use it (trigger phrases). See Anthropic’s guide for best practices.

​
Step 1: Create the Skill

  1. Click Skills in the left sidebar
  2. Click New Skill
  3. Enter a name (kebab-case, e.g. brand-voice) and description
  4. Click Create

​
Step 2: Add Files

  1. Open your new skill and create a SKILL.md file
  2. Add the YAML frontmatter with name and description
  3. Write your instructions below the frontmatter
  4. Optionally add files under references/ for deep reference content
Skill file editor

​
Step 3: Attach and Test

Attach the skill to an agent node (Skills field in the Model tab), open Chat, and ask something that should trigger it. You can debug triggering by asking the agent directly: “When would you use the brand-voice skill?”

​
Frequently Asked Questions

The agent reads the description field in SKILL.md’s YAML frontmatter. It uses this to decide whether the skill is relevant to the current task. Write descriptions that include specific trigger phrases, not just a general description.
Yes. The agent sees all attached skills and can use any combination of them within a single conversation.
Yes. Workflows reference skills by ID. Update the skill files and all agents using it pick up the changes immediately on the next run.
Yes. Open the skill settings and click Publish to add it to the public catalog. Others can then fork it into their own library.

​
What’s Next?

Connect Real Data

Connect external services and APIs to your agents via MCP

Multi-Agent Teams

Build networks of specialized agents, each with their own skills

Parallel Agents

Run multiple agents simultaneously for faster results

Deploy to Production

Take your workflow live with real users