What Are the Three Operating Modes of Humanizer?

Humanizer provides three distinct operating modes—Pasted Text (default), File, and Embedded—each tailored to different integration scenarios ranging from interactive editing to automated file processing.

The blader/humanizer skill rewrites AI-styled prose into natural-sounding text. Understanding the operating modes of Humanizer allows developers to choose the right integration strategy for their workflow, whether they are manually editing text blocks or automating content pipelines.

The Three Operating Modes of Humanizer

According to the specifications in SKILL.md, Humanizer supports three mutually exclusive ways to process and return rewritten content. Each mode is optimized for a specific use case and offers different output formats.

Pasted Text Mode (Default)

Pasted Text mode is the standard interaction when a user directly inputs a block of text for editing. As defined at line 48 in SKILL.md, this mode returns three components: the full draft of the rewrite, a short list of any remaining AI-style patterns detected, and the final cleaned-up text. This verbose output helps users understand exactly what changes the tool made and what patterns might still need manual review.

File Mode

File mode is designed for in-place document editing. When invoked via functions like humanizer.rewrite_file(), this mode writes only the final rewritten text back to the specified file path, as documented at line 50 in SKILL.md. Critically, File mode preserves structural elements including code blocks, inline code, shell commands, file paths, YAML metadata, data structures, and link targets. After writing the file, Humanizer returns a brief summary of what changed rather than the full text content.

Embedded Mode

Embedded mode returns only the final cleaned-up text with no additional metadata or draft stages. Specified at line 52 in SKILL.md, this mode is built for downstream automation scenarios such as generating pull-request titles, commit messages, or feeding rewritten content into other tools. Use this mode when Humanizer is called from scripts, CI pipelines, or other agent platforms where you need clean, machine-readable output.

How to Use Each Mode in Practice

The blader/humanizer repository exposes these modes through distinct function calls. Here are practical implementations for each operating mode:

# Pasted Text (default) mode - interactive editing

humanizer.rewrite(
    text="It's not just about the beat riding under the vocals; it's part of the aggression..."
)

# Returns: draft + remaining patterns + final text

# File mode - in-place file editing

humanizer.rewrite_file(
    path="docs/README.md"
)

# Rewrites the file in-place and returns a short change summary

# Embedded mode - automation and scripting

final_text = humanizer.rewrite_embedded(
    text="The new policy — announced without warning — affects thousands of workers."
)

# Returns: only the cleaned-up text string

Summary

• Pasted Text mode provides verbose feedback including drafts and pattern analysis, making it ideal for interactive editing sessions.
• File mode updates documents in-place while preserving code blocks and metadata, suitable for documentation pipelines.
• Embedded mode returns clean text only, optimized for automation tools and CI/CD workflows that consume rewritten content.

Frequently Asked Questions

What is the default operating mode in Humanizer?

Pasted Text mode is the default behavior when calling the standard humanizer.rewrite() function. This mode returns the full draft, a list of remaining patterns, and the final cleaned text, giving users complete visibility into the rewriting process.

Does File mode preserve code blocks and YAML metadata?

Yes. According to SKILL.md, File mode specifically leaves code blocks, inline code, commands, paths, YAML metadata, data structures, and link targets untouched while rewriting the prose content. This makes it safe for technical documentation and configuration files.

When should I use Embedded mode instead of Pasted Text mode?

Use Embedded mode when you need to pass the rewritten text directly to another tool or system without additional metadata. It is the optimal choice for generating commit messages, PR titles, or any automated pipeline where you need only the final cleaned-up string.

Where are the operating modes defined in the source code?

The three operating modes are formally specified in SKILL.md at lines 48, 50, and 52. Additional context about platform-specific packaging and agent integration can be found in AGENTS.md, while general usage guidance appears in README.md.

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