Where to Find the Source Code for Humanizer: A Complete Repository Guide

The complete source code for Humanizer lives in the SKILL.md file at the root of the blader/humanizer GitHub repository, with supporting configuration files in .claude-plugin/ and validation utilities in scripts/.

Humanizer is a pure-Markdown skill with no compilation step or binary artifacts. The entire logic—including 35 rewrite patterns, YAML metadata, and agent instructions—is defined in human-readable files that you can inspect, fork, or modify directly from the main branch.

The Core Skill Definition in SKILL.md

The primary source code for Humanizer resides in [SKILL.md](https://github.com/blader/humanizer/blob/main/SKILL.md). This single file implements all behavior and contains:

• YAML metadata defining the skill version (currently 2.11.2) and configuration parameters
• 35 rewrite patterns that power the text transformation logic
• Agent instructions that guide compatible platforms during execution

Unlike traditional software libraries, Humanizer requires no build step. The skill is ready to load directly by any agent platform that parses Markdown skill definitions, making the repository itself the complete distribution package.

Plugin Configuration Files

Humanizer distributes platform-specific manifests in the .claude-plugin/ directory to integrate with the Claude ecosystem.

.claude-plugin/plugin.json serves as the Claude plugin manifest, pointing agents to the core skill file using the "skill" key:

{
  "name": "humanizer",
  "skill": "SKILL.md",
  "metadata": {
    "version": "2.11.2"
  }
}

.claude-plugin/marketplace.json contains the marketplace descriptor for Claude Marketplace distribution, including metadata required for listing and version management.

Validation and Utility Scripts

The repository includes automation to ensure package integrity before publication.

scripts/validate-package.py performs pre-publish validation to verify:

• Metadata version consistency across SKILL.md, plugin.json, and README.md
• Pattern count alignment with the documented 35 rewrite rules
• File structure and required field validation

Run the validation script from the repository root:

python3 scripts/validate-package.py

An exit code of 0 indicates all skill files are consistent and the package is ready for deployment.

How to Access and Use the Source Code

You can consume the source code for Humanizer through several methods:

Direct integration for Claude developers: Clone the repository to access the raw skill files:

git clone https://github.com/blader/humanizer.git
cd humanizer

Programmatic consumption for custom agents: Fetch the skill directly for use with OpenAI-compatible tools or custom implementations:

import requests

skill_url = "https://raw.githubusercontent.com/blader/humanizer/main/SKILL.md"
skill_content = requests.get(skill_url).text

# Use the skill content as system instructions

import openai

response = openai.ChatCompletion.create(
    model="gpt-4o",
    messages=[
        {"role": "system", "content": skill_content},
        {"role": "user", "content": "Humanize this: The methodology utilized..."}
    ]
)

Because the source code is pure Markdown, you can load Humanizer into any environment capable of processing the SKILL.md instructions without compilation or dependency installation.

Summary

• The source code for Humanizer is contained entirely in the blader/humanizer repository with zero external dependencies.
• SKILL.md is the sole implementation file, containing YAML metadata and 35 rewrite patterns.
• .claude-plugin/plugin.json links the skill to Claude agents via the "skill": "SKILL.md" reference.
• scripts/validate-package.py ensures version consistency and validates package integrity before publication.
• The repository uses the main branch as the default and requires no build step for deployment.

Frequently Asked Questions

Where is the compiled binary for Humanizer?

There is no compiled binary. The source code for Humanizer is written entirely in Markdown and YAML, making it a zero-dependency skill that executes on any platform capable of parsing the SKILL.md file and following its instruction patterns.

What programming language is Humanizer written in?

The runtime logic is implemented in Markdown with YAML frontmatter. The only Python in the repository is the scripts/validate-package.py utility used for pre-publish validation, not for runtime functionality. The skill itself operates as a text-based instruction set for LLM agents.

How do I modify the rewrite patterns?

Edit the SKILL.md file directly in the repository root. This file contains all 35 rewrite patterns and agent behavior instructions. After modifying, execute python3 scripts/validate-package.py to verify that your changes maintain metadata consistency and pattern count alignment.

Can I use the source code without the Claude platform?

Yes. The SKILL.md file can be fetched via raw GitHub URLs and used by any OpenAI-compatible API or custom agent implementation. Simply include the file contents as system instructions to replicate Humanizer's functionality in environments outside the Claude ecosystem.

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