Allowed Types for Function Arguments in Claude Plugin Skill Files

Claude Plugin skill files support seven argument types: string, integer, float, boolean, enum, array, and object.

When building Claude Plugins, you'll define slash-command arguments in SKILL.md files using a lightweight JSON-style schema. Understanding the allowed function argument types is essential for creating valid skill definitions that Claude can parse and execute correctly. The anthropics/claude-plugins-community repository demonstrates these types through concrete skill implementations like tres-report-create and tres-upload-tx-header-validation.

Primitive Types in Skill Files

Skill files support four fundamental primitive types for arguments. These handle the most common data requirements in plugin commands.

string

The string type accepts arbitrary text input. Use it for names, descriptions, free-form content, or any text-based parameter.

arguments:
  - name: walletName
    type: string
    description: "Name of the wallet to create"

integer

The integer type represents whole signed numbers. It's ideal for counts, limits, indices, or any discrete numeric value.

arguments:
  - name: limit
    type: integer
    description: "Maximum number of rows to return"

float

The float type handles decimal numbers. Use it for prices, percentages, measurements, or any continuous numeric value.

arguments:
  - name: price
    type: float
    description: "Price per token"

boolean

The boolean type accepts true or false values. Common for toggle flags, dry-run modes, or enabling/disabling features.

arguments:
  - name: dryRun
    type: boolean
    default: false
    description: "Run the mutation without persisting changes"

Composite Types in Skill Files

Beyond primitives, skill files support three composite types that add structure and validation to arguments.

enum

The enum type restricts input to a fixed set of string values. This enforces valid choices and improves Claude's understanding of acceptable inputs.

arguments:
  - name: exportFormat
    type: enum
    values: [CSV, XLSX, JSON]
    description: "Desired file format for the generated report"

In tres-finance-plugin/skills/tres-report-create/SKILL.md, the enum type ensures only supported export formats are requested.

array

The array type represents a list of values with a single uniform element type. Specify the item type using the items field.

arguments:
  - name: tags
    type: array
    items: string
    description: "Optional tags for the asset"

The items field can reference any supported type, allowing arrays of strings, integers, or other primitives.

object

The object type accepts arbitrary JSON objects with nested fields. Use it for flexible metadata, configuration blocks, or complex structured data.

arguments:
  - name: metadata
    type: object
    description: "Optional free-form metadata"

Complete Skill File Example

The tres-report-create skill in tres-finance-plugin/skills/tres-report-create/SKILL.md demonstrates multiple argument types working together:

arguments:
  - name: reportName
    type: string
    description: "Human-readable name for the report to generate"
  - name: exportFormat
    type: enum
    values: [CSV, XLSX, JSON]
    description: "Desired file format for the generated report"
  - name: dryRun
    type: boolean
    default: false
    description: "Validate the request without actually creating the report"

This pattern—string for naming, enum for constrained choices, and boolean for flags—appears consistently across community plugins.

Type Mapping to Python

The skill file schema maps directly to Python type hints. Here's how the tres-report-create arguments translate:

from typing import Literal

def tres_report_create(
    reportName: str,
    exportFormat: Literal["CSV", "XLSX", "JSON"],
    dryRun: bool = False
) -> dict:
    """Generate a TRES Finance report."""
    # implementation …

    return {"status": "ok", "reportUrl": "..."}

Reference Files

These source files demonstrate argument types in production skill definitions:

• .claude-plugin/marketplace.json – Master manifest listing all community plugins and skills
• tres-finance-plugin/skills/tres-report-create/SKILL.md – Full arguments schema with string, enum, and boolean types
• tres-finance-plugin/skills/tres-upload-tx-header-validation/SKILL.md – Additional examples of string, enum, and boolean arguments

Summary

Claude Plugin skill files support seven function argument types across two categories:

• Primitives: string, integer, float, boolean — for basic data
• Composites: enum, array, object — for structured and validated data

Key implementation details from the anthropics/claude-plugins-community source:

• Arguments are declared under the arguments key in SKILL.md files
• enum requires a values array listing allowed strings
• array requires an items field specifying the element type
• boolean supports an optional default value

Frequently Asked Questions

Can I use nested objects inside array items?

No. The skill file schema in anthropics/claude-plugins-community supports single-type arrays only. For complex nested structures, use the object type and handle validation within your plugin implementation.

Is there a date or datetime type?

No. Neither date nor datetime appears as a dedicated type in community skill files. Represent temporal values as string arguments with clear format descriptions, or use integer for Unix timestamps.

What happens if I omit the type field?

Claude cannot parse the argument correctly. Every entry in the arguments array must include a valid type field. The parser in Claude's plugin system expects this schema for proper validation and tool-use generation.

Can I define custom validation beyond enum values?

No. The skill file schema provides structural types only. Implement additional validation logic—range checks, regex patterns, cross-field dependencies—in your plugin's handler code, not in the SKILL.md definition.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}