ai-linter 0.2.0

A validation tool for AI skills and agent configurations

AI Linter

• 1. Purpose
• 2. Features
• 3. Installation
  • 3.1. From PyPI (Coming Soon)
  • 3.2. From Source
  • 3.3. Development Installation
• 4. Command help
• 5. Usage
  • 5.1. Basic Usage
  • 5.2. Advanced Options
  • 5.3. Configuration File
• 6. Validation Rules
  • 6.1. Skills Validation
  • 6.2. Agents Validation
  • 6.3. Allowed Frontmatter Properties
  • 6.4. Pre-commit Integration
• 7. Exit Codes
  • 7.1. Development Workflow
• 8. Inspiration
• 9. Contributing
• 10. License

1. Purpose

AI Linter is a validation tool specifically designed for AI skills and agent configurations. It provides comprehensive linting and validation for:

• AI Skills: Validates SKILL.md files including frontmatter, content length, token count, and file references
• AI Agents: Validates AGENTS.md files and ensures proper structure without frontmatter
• File References: Checks that all referenced files exist and are accessible
• Content Quality: Enforces limits on content length and token counts to ensure optimal performance
• Project Structure: Validates overall project organization and file relationships

2. Features

• 🔍 Comprehensive Validation: Validates both skills and agents with detailed error reporting
• 📊 Token Counting: Uses tiktoken for accurate token count validation
• 📝 Frontmatter Parsing: Validates YAML frontmatter in skill files
• 🚫 File Reference Checking: Ensures all referenced files exist
• ⚙️ Configurable: Support for YAML configuration files
• 🎯 Selective Processing: Can process specific directories or auto-discover skills
• 📋 Detailed Logging: Multiple log levels with structured error reporting
• 🔧 Pre-commit Integration: Easy integration with pre-commit hooks

3. Installation

3.1. From PyPI (Coming Soon)

pip install ai-linter

3.2. From Source

# Clone the repository
git clone [e-mail]:fchastanet/ai-linter.git
cd ai-linter

# Install the package
pip install .

3.3. Development Installation

# Clone and install for development
git clone [e-mail]:fchastanet/ai-linter.git
cd ai-linter

# Install in development mode with all dev dependencies
pip install -e ".[dev]"

# Set up pre-commit hooks
pre-commit install

4. Command help

usage: aiLinter.py [-h]
          [--skills]
          [--max-warnings MAX_WARNINGS]
          [--ignore-dirs IGNORE_DIRS [IGNORE_DIRS ...]]
          [--log-level {ERROR,WARNING,INFO,DEBUG}]
          [--version]
          [--config-file CONFIG_FILE]
          directories [directories ...]

Quick validation script for skills

positional arguments:
  directories           Directories to validate

options:
  -h, --help            show this help message and exit
  --skills              Indicates that the input directories contain skills
  --max-warnings MAX_WARNINGS
                        Maximum number of warnings allowed before failing
  --ignore-dirs IGNORE_DIRS [IGNORE_DIRS ...]
                        List of directory patterns to ignore when validating AGENTS.md files
  --log-level {ERROR,WARNING,INFO,DEBUG}
                        Set the logging level
  --version             Show the version of the AI Linter
  --config-file CONFIG_FILE
                        Path to the AI Linter configuration file

5. Usage

After installation, you can use AI Linter via the ai-linter command or directly with Python.

5.1. Basic Usage

# Using the console script (recommended)
ai-linter /path/to/directory

# Auto-discover and validate all skills in a directory
ai-linter --skills /path/to/skills/directory

# Using the Python module directly
python src/aiLinter.py --skills examples/

5.2. Advanced Options

# Set maximum allowed warnings
ai-linter --max-warnings 5 /path/to/directory

# Ignore specific directories
ai-linter --ignore-dirs node_modules build /path/to/directory

# Set log level
ai-linter --log-level DEBUG /path/to/directory

# Use custom config file
ai-linter --config-file custom-config.yaml /path/to/directory

# Show version
ai-linter --version

5.3. Configuration File

Create a .ai-linter-config.yaml file in your project root:

# Logging configuration
log_level: INFO  # DEBUG, INFO, WARNING, ERROR

# Maximum warnings before failing
max_warnings: 10

# Directories to ignore during validation
ignore_dirs:
  - .git
  - __pycache__
  - node_modules
  - build
  - dist

6. Validation Rules

6.1. Skills Validation

• ✅ SKILL.md file must exist
• ✅ Valid YAML frontmatter with required properties
• ✅ Content length must not exceed 500 lines
• ✅ Token count must not exceed 5000 tokens
• ✅ All file references must exist and be accessible
• ✅ Frontmatter must contain valid metadata

6.2. Agents Validation

• ✅ AGENTS.md file structure validation
• ✅ No frontmatter allowed in agent files
• ✅ Content length must not exceed 500 lines
• ✅ Token count must not exceed 5000 tokens
• ✅ All file references must be valid

6.3. Allowed Frontmatter Properties

name: string
description: string
license: string
allowed-tools: array
metadata: object
compatibility: object

6.4. Pre-commit Integration

• ✅ .pre-commit-hooks.yaml - Hook definitions for external use
• ✅ .pre-commit-config.yaml - Local development configuration with:
  • AI Linter validation
  • Black code formatting
  • isort import sorting
  • flake8 linting
  • mypy type checking
  • bandit security scanning
  • YAML/Markdown validation

See .pre-commit-hooks.yaml for ready-to-use pre-commit hooks. Add to your .pre-commit-config.yaml:

repos:
  - repo: https://github.com/fchastanet/ai-linter
    rev: v0.1.0
    hooks:
      - id: ai-linter-skills
        args: [--max-warnings, '5']

7. Exit Codes

• 0: Success (no errors, warnings within limits)
• 1: Failure (errors found or too many warnings)

7.1. Development Workflow

# Setup development environment
make install-dev

# Run all checks
make check-all

# Test the linter
make validate

8. Inspiration

This tool was inspired by Anthropic's skill validation script and adapted for broader AI development workflows.

9. Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

10. License

This project is licensed under the MIT License - see the LICENSE file for details.