Configuration Reference

Everything you can put in your ai-rulez.yml file.

Main Structure

The key sections in your config file.

metadata

Basic info about your project.

metadata:
  name: "My Project"
  version: "1.0.0"
  description: "A brief description of the project."

presets

The recommended way to configure output files for common AI tools. This is a list of strings, where each string is a supported preset name.

# Use presets for popular tools
presets:
  - "popular" # Includes Claude, Cursor, Windsurf, and Copilot
  - "gemini"

See the "Configuring Outputs" section below for a full list of available presets.

outputs

The advanced method for configuring output files. This gives you fine-grained control over file paths, naming schemes, and templates. Use this when a preset doesn't meet your needs or you have a custom tool.

# For advanced, custom configurations
outputs:
  - path: "CUSTOM_PROMPT.md"
    template:
      type: "inline"
      value: "My custom prompt template..."

extends

A string specifying a single base configuration file to inherit from. The path can be a local file or a remote URL.

includes

A list of strings specifying other ai-rulez.yml files to merge into your configuration. Paths can be local files or remote URLs.

rules

A list of rule objects that define instructions for your AI assistant. See the rule object reference below for details.

sections

A list of section objects that define large blocks of documentation or context. See the section object reference below for details.

agents

A list of agent objects that define specialized sub-assistants. See the agent object reference below for details.

mcp_servers

A list of MCP server objects for integrating external tools. See the mcp_server object reference below for details.

commands

A list of custom slash command objects for your AI assistant. See the command object reference below for details.

gitignore

Controls whether ai-rulez automatically updates your .gitignore file with generated files.

gitignore: false  # Disable automatic .gitignore updates

By default, ai-rulez will add all generated files to your .gitignore to prevent them from being committed to version control. This is usually what you want, since these files are regenerated from your configuration.

• Default: true (automatically update .gitignore)
• Set to false: If you want to manage .gitignore manually or commit generated files

Configuring Outputs: Presets vs. Outputs

ai-rulez offers two ways to define the files it generates.

Using Presets (Recommended)

For most users, presets are the simplest and recommended method. They are a shorthand for the most common AI tool configurations.

presets:
  - "popular"
  - "gemini"

Available Presets:

• popular: A convenient bundle that includes claude, cursor, windsurf, and copilot.
• claude: Generates CLAUDE.md for Anthropic Claude.
• cursor: Generates rules for Cursor in .cursor/rules/.
• copilot: Generates .github/copilot-instructions.md for GitHub Copilot.
• gemini: Generates GEMINI.md for Google Gemini.
• windsurf: Generates rules for Windsurf in .windsurf/.
• continue: Generates rules and prompts for Continue.dev.
• cline: Generates rules for Cline in .clinerules/.
• amp / codex: Generates AGENTS.md for Sourcegraph tools.

Using Outputs (Advanced)

If you need more control than presets offer, use the outputs key. This is useful for:

• Generating files for custom or unsupported tools.
• Using custom file paths or naming schemes.
• Applying custom templates to the generated files.

See the output Object reference below for all available options.

Combining Presets and Outputs

You can use both presets and outputs in the same configuration file. This is useful when you want to use a standard preset but need to add or modify one specific output.

When both are used:

1. The outputs from the presets are generated first.
2. The outputs list is then added.
3. If there is a path conflict (e.g., you use the claude preset but also define an output with the path CLAUDE.md), the entry from the outputs list will take precedence.

# Use the popular preset, but override the Claude file with a custom template
presets:
  - "popular"

outputs:
  - path: "CLAUDE.md" # This will override the default Claude output from the "popular" preset
    template:
      type: "file"
      value: "./my-custom-claude-template.tmpl"

Supporting Custom Tools

What if your favorite tool isn't on the preset list? No problem. ai-rulez is designed to be extensible to any tool that consumes text-based configuration files.

The key is the outputs list. By defining an output, you can tell ai-rulez to generate a file at any path with any content you need. This is where the power of templates comes in.

Example: Supporting a Custom Linter

Imagine you use a custom linter called "CodeSniffer" that reads its configuration from a .codesniffer.json file. You want to provide it with a list of important directories to scan, which you've defined in your rules.

Here’s how you would configure that in ai-rulez.yml:

rules:
  - name: "Important Directories"
    content: "src/"
  - name: "Other Directories"
    content: "pkg/"

outputs:
  - path: ".codesniffer.json"
    template:
      type: "inline"
      value: |
        {
          "scan_directories": [
            {{- range $i, $rule := .Rules -}}
              {{- if $i -}},{{- end -}}
              "{{- .Content -}}"
            {{- end -}}
          ]
        }

This configuration would generate a .codesniffer.json file with the following content:

{
  "scan_directories": [
    "src/",
    "pkg/"
  ]
}

This demonstrates how you can use the outputs key and Go's templating engine to generate configuration for any tool, in any format.

Object Reference

This section details the fields for each type of object in the configuration.

output Object

Defines a file or directory to be generated. This is the advanced alternative to using presets.

• path (string, required): The output path. If it ends with a /, it is treated as a directory.
• type (string): For directory outputs, specifies the content type to generate (rule or agent).
• naming_scheme (string): For directory outputs, a pattern for filenames (e.g., {name}.md).
• template (object): A structured object to define the template.
• type (string, required): builtin, file, or inline.
• value (string, required): The name, path, or content of the template.

rule Object

• name (string, required): The name of the rule.
• content (string, required): The content of the rule.
• id (string): An optional unique identifier.
• priority (string): critical, high, medium, low, or minimal.
• targets (array of strings): A list of output paths this rule applies to.

section Object

• name (string, required): The name of the section.
• content (string, required): The content of the section.
• id (string): An optional unique identifier for overriding.
• priority (string): critical, high, medium, low, or minimal.
• targets (array of strings): A list of output paths this section applies to.

agent Object

• name (string, required): The name of the agent.
• description (string, required): A description of the agent's purpose.
• id (string): An optional unique identifier for overriding.
• priority (string): The agent's priority.
• tools (array of strings): A list of tools available to the agent.
• model (string): An optional model identifier (e.g., claude-3-opus-20240229, gpt-4).
• system_prompt (string): The system prompt for the agent.
• template (object): A structured template object for the system prompt.
• type (string, required): builtin, file, or inline.
• value (string, required): The template content, file path, or builtin name.
• targets (array of strings): A list of output paths this agent applies to.

mcp_server Object

• name (string, required): The name of the server.
• id (string): An optional unique identifier for overriding.
• description (string): A description of the server.
• command (string): The command to start the server.
• args (array of strings): Arguments for the command.
• env (object): Environment variables for the server.
• transport (string): stdio, sse, or http.
• url (string): The URL for remote servers.
• enabled (boolean): Whether the server is enabled.
• targets (array of strings): A list of output paths this server applies to.

command Object

• name (string, required): The name of the command (e.g., new-task).
• description (string, required): A description of the command.
• id (string): An optional unique identifier for overriding.
• aliases (array of strings): Alternative names for the command.
• usage (string): A usage example (e.g., /new-task <description>).
• system_prompt (string): The system prompt to use for this command.
• shortcut (string): A keyboard shortcut.
• enabled (boolean): Whether the command is enabled.
• targets (array of strings): A list of output paths this command applies to.