Why must scripts accept an output path?

To allow callers to specify where results go, keeping the skill directory focused on code while output destinations are configurable.

How do I learn more about a skill’s usage?

Use skills_read(path='skills/ /SKILL.md') to read the main guide, or skills_read(path='skills/ /docs/..') to view detailed documentation.

skill-creator

Scanned

npx machina-cli add skill nemori-ai/agent_skills/skill-creator --openclaw

Files (1)

SKILL.md

3.2 KB

Skill Creator

This skill teaches you how to create Skill Packages—complete directory structures containing guides, scripts, and data.

⚠️ Path Guidelines

Rule: Scripts should accept output path parameters, letting callers specify output locations. Skill directories only store code.

For detailed guidelines: skills_read(path="skills/skill-creator/docs/script-guidelines.md")

What is a Skill Package?

A skill package is not just a Markdown file, but a complete directory structure:

skill-name/
├── SKILL.md          # Entry guide (required)
├── scripts/          # Executable scripts
│   ├── main.py
│   └── pyproject.toml  (dependency config)
├── data/             # Templates and data files
└── docs/             # Detailed documentation (optional)

📖 Choose Guide by Scenario

Based on your needs, read the corresponding detailed documentation:

🆕 Create New Skill from Scratch

First time creating a skill package, learn the complete process (5-step creation method).

skills_read(path="skills/skill-creator/docs/quick-start.md")

📝 Write/Modify SKILL.md Documentation

Learn the structure and template for SKILL.md.

skills_read(path="skills/skill-creator/docs/skillmd-template.md")

🔧 Add Scripts to Skill

Learn Python/Bash script writing standards and dependency management.

skills_read(path="skills/skill-creator/docs/script-guidelines.md")

🔄 Script Debugging Failed, Need Iteration

How to properly clean up old versions after creating replacement scripts.

skills_read(path="skills/skill-creator/docs/iteration-and-cleanup.md")

📚 View Complete Example

Reference a complete skill creation process (code reviewer skill).

skills_read(path="skills/skill-creator/docs/full-example.md")

Command Quick Reference

Tool	Purpose	Example
`skills_ls(path="skills")`	List all skills	View available skills
`skills_ls(path="skills/<name>")`	List files in skill	Check file structure
`skills_read(path="skills/<name>/SKILL.md")`	Read skill documentation	Learn skill usage
`skills_create(name, description, instructions)`	Create skill skeleton	Create new skill
`skills_write(path, content)`	Add/overwrite file	Add script
`skills_run(name, command)`	Execute skill command	Test script
`skills_bash(command, cwd)`	Execute shell command	Delete/rename file

Quick Start

Creating the simplest skill requires only 2 steps:

# 1. Create skill skeleton
skills_create(
    name="hello-world",
    description="Example skill",
    instructions="# Hello World\n\nRun `skills_run(name=\"hello-world\", command=\"python scripts/hello.py\")`"
)

# 2. Add script
skills_write(
    path="skills/hello-world/scripts/hello.py",
    content='print("Hello, World!")'
)

# 3. Test
skills_run(name="hello-world", command="python scripts/hello.py")

Need more detailed guidance? Read Quick Start Guide.

Source

git clone https://github.com/nemori-ai/agent_skills/blob/main/agent_skills/skills/skill-creator/SKILL.mdView on GitHub

Overview

Skill Packages are complete directories that bundle guides, executable scripts, data templates, and optional docs. The entry point is SKILL.md, while scripts should be path-parameter aware and stored separately in scripts/. This structure keeps code, data, and documentation organized for easy reuse and sharing.

How This Skill Works

A skill package lives in a dedicated directory (skill-name/) containing SKILL.md, a scripts/ folder for executables, a data/ folder for templates, and an optional docs/ folder for in-depth guidance. Scripts must accept an output path, ensuring callers can specify where results go. You interact with the package via helper commands like skills_create, skills_write, skills_run, and skills_read to build, populate, and test the package.

When to Use It

You need to create a new, fully structured skill package from scratch with scripts, templates, and documentation.
You want to modify SKILL.md structure or templates to align with a standardized format.
You are adding or updating executable scripts and want to manage dependencies in a dedicated scripts/ folder.
You need to clean up or replace old script versions while preserving a clear directory structure.
You want to review a complete example of a skill creation process to guide implementation.

Quick Start

Step 1: Create a skill skeleton with skills_create(name='hello-world', description='Example skill', instructions='...')
Step 2: Add script to skills/hello-world/scripts/hello.py that prints a message using skills_write
Step 3: Test with skills_run(name='hello-world', command='python scripts/hello.py')

Best Practices

Place SKILL.md as the entry guide inside the skill directory and keep code separated in scripts/.
Ensure all scripts accept an explicit output path parameter to support flexible destinations.
Store templates and data files under data/ and keep dependencies described in a dedicated setup (e.g., pyproject.toml) when applicable.
Follow the 5-step creation method for new skills and consult the Quick Start guidance for consistency.
Use the optional docs/ folder for detailed, long-form documentation and read via skills_read when needed.

Example Use Cases

hello-world: a minimal skeleton created with a script, a hello.py, and a test run showing the basics.
image-processor: a data-heavy skill with templates in data/ and image-processing scripts in scripts/.
code-reviewer (complete skill): a full example referenced in the complete skill creation process.
report-generator: outputs artifacts to a user-specified path via script parameters.
documentation-focused SKILL.md modification: updating the SKILL.md structure and templates for clarity.

Frequently Asked Questions

Add this skill to your agents