Use Skilo

Use Skilo

Skilo is a CLI tool for developing Agent Skills.

Installation

curl -sSfL https://raw.githubusercontent.com/manuelmauro/skilo/main/install.sh | sh

Quick Start

skilo new my-skill        # Create from template
skilo add owner/repo      # Install from git
skilo lint .              # Validate against spec
skilo fmt .               # Format SKILL.md files

Run skilo -h for all commands and options.

ALWAYS Validate Skills

Run skilo lint and skilo fmt before committing any skill changes. These commands enforce the Agent Skills Specification and catch common issues.

# Validate and format in one command
skilo check .

# Strict mode for CI (warnings become errors)
skilo check --strict .

Skill Structure

my-skill/
├── SKILL.md        # Required: manifest with YAML frontmatter
├── scripts/        # Optional: executable code
├── references/     # Optional: additional docs (loaded on-demand)
└── assets/         # Optional: static resources

Frontmatter Requirements

Required Fields

name (1-64 characters)

• Lowercase alphanumeric and hyphens only (a-z, 0-9, -)
• Must NOT start/end with - or contain --
• Must match parent directory name

description (1-1024 characters)

• Describe WHAT the skill does AND WHEN to use it
• Include keywords that help agents match user requests

Optional Fields

• license - Keep short, reference LICENSE file for details
• compatibility (max 500 chars) - Environment requirements (e.g., Requires git, docker)
• metadata - Key-value pairs for custom properties
• allowed-tools - Space-delimited list of pre-approved tools (experimental)

Best Practices

Write Effective Descriptions

The description is how agents decide when to activate your skill. Be specific.

# Good - specific, keyword-rich
description: Extracts text and tables from PDF files, fills PDF forms, and merges multiple PDFs. Use when working with PDF documents or when the user mentions PDFs, forms, or document extraction.

# Bad - vague, no activation cues
description: Helps with PDFs.

Keep Body Content Concise

The entire SKILL.md body loads when the skill activates. Keep it under 500 lines.

• Include step-by-step instructions
• Add examples of inputs and outputs
• Cover common edge cases
• Move detailed material to references/

Use Progressive Disclosure

Structure skills to minimize context usage:

1. Metadata (~100 tokens) - name + description loaded at startup for all skills
2. Instructions (<5000 tokens) - Full body loaded when skill activates
3. Resources (on-demand) - scripts/, references/, assets/ loaded only when needed

Write Self-Contained Scripts

Scripts in scripts/ should:

• Be self-contained or clearly document dependencies
• Include helpful error messages
• Handle edge cases gracefully

Organize References Effectively

Files in references/ are loaded on-demand:

• Keep files focused (one file = one concept)
• Use relative paths from skill root
• Avoid deeply nested reference chains

Lint Rules

Skilo enforces these rules (configure in .skilorc.toml):

CI Integration

- name: Validate skills
  run: |
    curl -sSfL https://raw.githubusercontent.com/manuelmauro/skilo/main/install.sh | sh
    skilo check --strict skills/