Skip to content
skillx

Hello World

The hello-world skill is the simplest possible skillx skill. It demonstrates the full lifecycle: scan, run, observe, cleanup.

SKILL.md

Here is the complete content of examples/skills/hello-world/SKILL.md:

---
name: hello-world
description: A simple greeting skill that demonstrates skillx capabilities
author: skillx-run
version: "1.0.0"
license: MIT
tags:
  - example
  - starter
---


# Hello World Skill


You are an assistant using the skillx Hello World skill.


## Instructions


When invoked, greet the user warmly and explain what skillx skills are:


1. **skillx** lets you run any agent skill without installing it
2. **Skills** are reusable instruction sets that enhance your AI coding agent
3. This skill was injected via `skillx run` — it will be automatically cleaned up when the session ends


Then help the user with whatever task they describe.


## Example Interactions


- "hello" → Greet and briefly introduce skillx
- "help me write a function" → Greet, then assist with coding
- "what can you do?" → Explain your capabilities enhanced by this skill

Structure

This is a single-file skill — just a SKILL.md with no scripts or references:

hello-world/
└── SKILL.md

The frontmatter provides metadata (name, description, author, version, license, tags). The body contains the instructions the agent will follow.

Step 1: Scan

You can use the GitHub source directly from the docs:

Terminal window
skillx scan github:skillx-run/skillx/examples/skills/hello-world

If you are already inside a local clone of this repository, the equivalent local path is:

Terminal window
skillx scan ./examples/skills/hello-world

Expected output:

  Scanning  hello-world
  ────────────────────────────────
  ✓ PASS — No issues found


  Files scanned: 1
  Risk level: PASS

The skill passes cleanly because it contains no dangerous patterns — no external URLs, no shell commands, no sensitive directory references.

Step 2: Run

Run the skill directly from GitHub:

Terminal window
skillx run github:skillx-run/skillx/examples/skills/hello-world "Hello, what can you do?"

If you are studying the repository locally, use:

Terminal window
skillx run ./examples/skills/hello-world "Hello, what can you do?"

skillx performs the full lifecycle:

  1. Resolve — Resolves the source, whether it comes from GitHub or a local path
  2. Scan — Checks for security issues (PASS)
  3. Detect agent — Identifies your active AI coding agent
  4. Inject — Copies the skill into the agent’s instruction directory
  5. Wait — Keeps the session active while you interact with the agent
  6. Cleanup — Removes all injected files when you press Ctrl+C

Step 3: Observe

Once injected, your agent has access to the skill’s instructions. The agent will greet you and explain what skillx is, then help with your request.

When you are done, press Ctrl+C to end the session. skillx removes the injected files and archives the session manifest.

Running with Different Agents

By default, skillx auto-detects your active agent. You can override this with --agent:

Claude Code

Terminal window
skillx run github:skillx-run/skillx/examples/skills/hello-world --agent claude-code "Hello"

Injects into Claude Code’s CLAUDE.md instruction file.

Codex

Terminal window
skillx run github:skillx-run/skillx/examples/skills/hello-world --agent codex "Hello"

Injects into Codex’s instruction directory.

Cursor

Terminal window
skillx run github:skillx-run/skillx/examples/skills/hello-world --agent cursor "Hello"

Injects into Cursor’s .cursor/rules/ directory.

Universal (Fallback)

Terminal window
skillx run github:skillx-run/skillx/examples/skills/hello-world --agent universal "Hello"

Copies the skill to the project root and displays the content for manual use with any agent.

skillx fetches the GitHub source, caches it locally, scans it, and injects it — all in one command.

If you are testing these same flows from a local clone of skillx-run/skillx, replace the GitHub source with ./examples/skills/hello-world.

Why this example exists

This example exists as the smallest complete skill in the repository, so you can see the minimum viable SKILL.md shape before you study any of the more opinionated patterns.

Next Steps

If you want to see a more practical single-file skill, open Code Review. If you want to turn this minimal shape into your own skill, continue with Writing Skills. If you are ready to skip templates and use something mature from GitHub, go to Famous Skills.