How Humanizer Handles File Mode Processing: A Complete Guide

Humanizer processes files by running its four-step rewrite pipeline exclusively on natural-language prose while preserving all code blocks, YAML front matter, and structured data.

Humanizer is an open-source skill designed to transform AI-generated text into natural, human-like prose. When you invoke Humanizer file mode, the skill targets a specific file path on disk and executes its rewrite pipeline while strictly protecting technical elements like code fences and configuration data.

Understanding Humanizer File Mode

In SKILL.md (line 48), the file mode specification defines how the blader/humanizer repository handles on-disk documents. Rather than processing raw text indiscriminately, the skill applies its transformation logic only to natural-language sections, leaving all structured content intact.

The Four-Step Rewrite Pipeline

When you point Humanizer at a file, it executes this workflow:

1. Mark tells: Identifies AI-style linguistic patterns and markers in the prose.
2. Draft rewrite: Generates humanized alternatives for the flagged sections.
3. Check draft: Validates the rewritten content for quality and consistency.
4. Final version: Produces the polished prose for file writeback.

Crucially, the skill only writes the final prose back to the file. All preserved technical elements remain untouched in their original positions according to the source specification.

Content Preservation Rules

Humanizer explicitly safeguards the following elements during file mode processing:

• Code blocks and inline code within backticks
• Command snippets and shell instructions
• File-system paths and directory references
• YAML front-matter headers
• JSON and YAML data sections
• Link targets and URL structures

This design ensures that running Humanizer on documentation or markdown files poses zero risk to executable code, configuration syntax, or hyperlink integrity.

How to Use Humanizer File Mode

The repository provides two primary interfaces for invoking file mode, documented in README.md (line 44).

Command-Line Invocation via Skills CLI

You can invoke Humanizer globally using the Skills CLI:

npx skills add blader/humanizer --global
/humanizer
Humanize the prose in docs/launch-post.md

The command loads docs/launch-post.md, processes only the natural-language sections, and overwrites the file with the humanized version while preserving all code fences and metadata.

Chat Interface Usage

Alternatively, invoke the skill directly within a chat interface:

/humanizer
Humanize the prose in docs/launch-post.md

This method triggers the same four-step pipeline:

1. Loads the specified file path
2. Detects AI-style tells exclusively in natural-language sections
3. Rewrites prose while maintaining every code block, inline command, URL, and YAML block
4. Overwrites the original file and outputs a brief change summary

Technical Implementation Details

The behavior is governed by configuration files in the repository root.

Source Files and Configuration

• SKILL.md: Contains the core skill definition and the file mode specification at line 48. This document defines the content preservation boundaries and the rewrite pipeline steps.
• README.md: Provides user-facing documentation and usage examples at line 44, demonstrating both CLI and chat-based invocation methods.
• AGENTS.md: Stores metadata about agent packaging and distribution, though it does not directly affect the file mode processing logic.

When processing completes, Humanizer returns a concise summary describing what content was modified, allowing you to verify the scope of changes without reviewing the entire file diff.

Summary

• Humanizer file mode targets specific disk paths and runs a four-step pipeline (mark tells → draft rewrite → check draft → final version) exclusively on natural language.
• The skill preserves all technical content including code blocks, inline code, YAML front matter, JSON data sections, file paths, and link targets.
• Configuration and behavior are defined in SKILL.md (line 48) with usage examples in README.md (line 44).
• Users can invoke file mode via the Skills CLI (npx skills add blader/humanizer --global) or through chat interfaces using the /humanizer command.
• After processing, the skill overwrites the original file with humanized prose and returns a summary of changes.

Frequently Asked Questions

What is Humanizer file mode?

Humanizer file mode is a specialized operation that processes text files on disk by applying the skill’s rewrite pipeline only to natural-language prose. It preserves all code blocks, YAML front matter, and structured data while updating the file in place and returning a summary of modifications.

Which file types does Humanizer support in file mode?

According to the SKILL.md specification in the blader/humanizer repository, Humanizer handles any text-based file including markdown documentation, configuration files with YAML front matter, and mixed-format documents. The skill specifically safeguards code fences, inline commands, file paths, and data sections regardless of file extension.

Does Humanizer modify code blocks when processing files?

No. The file mode implementation explicitly preserves code blocks, inline code snippets, command examples, and file-system paths as defined in the source specification. Only natural-language prose sections receive humanization treatment, ensuring technical accuracy remains intact.

How do I check what Humanizer changed in my file?

After processing completes, Humanizer outputs a brief summary of modifications made to the file. This summary describes which sections were humanized without requiring you to parse the entire file diff, providing immediate visibility into the scope of changes applied during the file mode operation.

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