# How to Manually Install the Humanizer Skill for LLM Agents > Learn how to manually install the Humanizer skill for your LLM agent. Follow simple steps to copy the SKILL.md file from the blader/humanizer repository into your agent's skill directory. - Repository: [Siqi Chen/humanizer](https://github.com/blader/humanizer) - Tags: how-to-guide - Published: 2026-09-06 --- **Manually installing the Humanizer skill requires copying the [`SKILL.md`](https://github.com/blader/humanizer/blob/main/SKILL.md) file from the blader/humanizer repository into your LLM agent's skill directory and optionally adding Claude-specific plugin descriptors if using Claude Desktop.** The Humanizer skill is a pure-Markdown utility that contains only data—metadata, rewrite patterns, and usage instructions—requiring no code compilation or runtime dependencies. Because it works entirely through the agent's skill-loader, you can manually install it on any compatible LLM agent by simply transferring the skill definition file. ## Understanding the Humanizer Skill Structure The blader/humanizer repository follows a data-only architecture that makes manual installation straightforward. ### Pure Markdown Design Unlike traditional software packages, Humanizer contains no binaries, Python modules, or external service dependencies. The entire skill is defined in a single [`SKILL.md`](https://github.com/blader/humanizer/blob/main/SKILL.md) file that the agent's skill-loader parses directly. This design means you do not need package managers, build tools, or language runtimes to manually install the Humanizer skill. ### Metadata and Rewrite Patterns The [`SKILL.md`](https://github.com/blader/humanizer/blob/main/SKILL.md) file stores **metadata** at the top using `metadata.version`, allowing agents to recognize the skill version without separate package files. The file contains **35 rewrite patterns** sourced from Wikipedia's *Signs of AI writing*, organized into five categories: - Content patterns - Language-grammar patterns - Style patterns - Chatbot patterns - Filler patterns ### Embedded Usage Instructions The skill includes natural language usage instructions that teach agents how to invoke it via the `/humanizer` command or plain-language prompts. Because these instructions are embedded directly in [`SKILL.md`](https://github.com/blader/humanizer/blob/main/SKILL.md), the skill is self-documenting upon installation. ## Prerequisites for Manual Installation Before you manually install the Humanizer skill, ensure you have: - Access to download files from the blader/humanizer repository - Write permissions to your LLM agent's skill directory - A compatible agent that supports Markdown-based skills (Claude Desktop, OpenAI Code, or similar) ## Step-by-Step Manual Installation Guide Follow these steps to manually install the skill without using the Skills CLI. ### 1. Download the SKILL.md File First, obtain the core skill definition from the repository. You can download it directly using `curl` or through your browser: ```bash curl -L https://github.com/blader/humanizer/raw/main/SKILL.md -o SKILL.md ``` The file must retain the exact name [`SKILL.md`](https://github.com/blader/humanizer/blob/main/SKILL.md) because the agent's skill-loader specifically looks for this filename when scanning directories. ### 2. Locate Your Agent's Skill Directory The exact path depends on which LLM agent you are using: - **Claude Desktop**: `/skills/` or `~/.claude/skills/` - **OpenAI Code**: `/skills/` - **Other agents**: Consult your agent's documentation for the "skill directory" or "plugins" folder ### 3. Copy SKILL.md to the Skill Folder Place the downloaded file directly inside the agent's skill folder. Do not rename the file or place it in subdirectories unless your agent specifically supports nested skill structures. ```bash mkdir -p "$AGENT_SKILL_DIR" cp SKILL.md "$AGENT_SKILL_DIR/" ``` ### 4. Optional: Add Claude Plugin Descriptors If you are using **Claude Desktop** and want the skill to be discoverable through Claude's plugin system, you must also copy two JSON descriptor files: - [`.claude-plugin/plugin.json`](https://github.com/blader/humanizer/blob/main/.claude-plugin/plugin.json) - [`.claude-plugin/marketplace.json`](https://github.com/blader/humanizer/blob/main/.claude-plugin/marketplace.json) Download these files from the repository and place them in the same skill directory: ```bash curl -L https://github.com/blader/humanizer/raw/main/.claude-plugin/plugin.json \ -o "$AGENT_SKILL_DIR/.claude-plugin/plugin.json" curl -L https://github.com/blader/humanizer/raw/main/.claude-plugin/marketplace.json \ -o "$AGENT_SKILL_DIR/.claude-plugin/marketplace.json" ``` These files are optional for other agents but required for full Claude Desktop integration. ### 5. Reload the Agent Restart or reload your LLM agent so it rescans its skill directory and registers the new skill. After reloading, the Humanizer skill is available immediately via the `/humanizer` command or natural-language requests. ```bash # Example for Claude Desktop on macOS pkill -f claude && open -a "Claude Desktop" ``` ## Complete Installation Script For convenience, here is the complete command sequence to manually install the Humanizer skill with optional Claude plugin support: ```bash # Define the skill directory (adjust for your agent) AGENT_SKILL_DIR="$HOME/.claude/skills" # Create directory and download core skill file mkdir -p "$AGENT_SKILL_DIR" curl -L https://github.com/blader/humanizer/raw/main/SKILL.md \ -o "$AGENT_SKILL_DIR/SKILL.md" # Optional: Download Claude plugin descriptors mkdir -p "$AGENT_SKILL_DIR/.claude-plugin" curl -L https://github.com/blader/humanizer/raw/main/.claude-plugin/plugin.json \ -o "$AGENT_SKILL_DIR/.claude-plugin/plugin.json" curl -L https://github.com/blader/humanizer/raw/main/.claude-plugin/marketplace.json \ -o "$AGENT_SKILL_DIR/.claude-plugin/marketplace.json" # Restart agent (Claude Desktop example) pkill -f claude && open -a "Claude Desktop" ``` ## Key Files in the Repository Understanding these files helps troubleshoot manual installations: - **[`SKILL.md`](https://github.com/blader/humanizer/blob/main/SKILL.md)**: The core skill definition containing metadata, the 35 rewrite patterns, and usage instructions. This is the only required file for basic installation. - **[`.claude-plugin/plugin.json`](https://github.com/blader/humanizer/blob/main/.claude-plugin/plugin.json)**: Claude Desktop plugin descriptor that registers the skill with Claude's plugin system. - **[`.claude-plugin/marketplace.json`](https://github.com/blader/humanizer/blob/main/.claude-plugin/marketplace.json)**: Marketplace metadata for Claude plugin discovery. - **[`AGENTS.md`](https://github.com/blader/humanizer/blob/main/AGENTS.md)**: Guidelines for maintaining compatibility across different LLM agents. - **[`README.md`](https://github.com/blader/humanizer/blob/main/README.md)**: Human-readable documentation including alternative installation methods. ## Summary - **Manually installing the Humanizer skill** requires only copying [`SKILL.md`](https://github.com/blader/humanizer/blob/main/SKILL.md) into your agent's skill directory. - The skill is **pure Markdown** with no runtime dependencies, binaries, or build steps required. - The [`SKILL.md`](https://github.com/blader/humanizer/blob/main/SKILL.md) file contains **metadata.version** at the top for version recognition and **35 rewrite patterns** organized into content, grammar, style, chatbot, and filler categories. - **Claude Desktop users** should additionally copy [`.claude-plugin/plugin.json`](https://github.com/blader/humanizer/blob/main/.claude-plugin/plugin.json) and [`.claude-plugin/marketplace.json`](https://github.com/blader/humanizer/blob/main/.claude-plugin/marketplace.json) for full plugin integration. - After copying files, **reload the agent** to activate the skill, which becomes available via `/humanizer` or natural language prompts. ## Frequently Asked Questions ### Does the Humanizer skill require Python or other runtime dependencies? No. The Humanizer skill is a pure-Markdown file ([`SKILL.md`](https://github.com/blader/humanizer/blob/main/SKILL.md)) that contains only data—metadata, patterns, and usage instructions. It works entirely through the agent's skill-loader and does not require Python, Node.js, or any external services to function. ### What is the difference between SKILL.md and the .claude-plugin JSON files? [`SKILL.md`](https://github.com/blader/humanizer/blob/main/SKILL.md) is the **core skill definition** required by all agents; it contains the 35 rewrite patterns and invocation instructions. The `.claude-plugin/*.json` files are **optional metadata descriptors** specifically for Claude Desktop that enable plugin discovery and marketplace integration. If you are not using Claude Desktop, you only need [`SKILL.md`](https://github.com/blader/humanizer/blob/main/SKILL.md). ### Can I manually install Humanizer on agents other than Claude Desktop? Yes. Because the skill uses a data-only architecture, you can manually install it on any LLM agent that supports Markdown-based skills, including OpenAI Code and other compatible platforms. Simply copy [`SKILL.md`](https://github.com/blader/humanizer/blob/main/SKILL.md) into the agent's designated skill directory and reload. ### How do I verify that the skill is properly installed? After reloading your agent, test the installation by issuing the `/humanizer` command or asking the agent to "humanize this text." If the skill is properly loaded, the agent will recognize the command and apply the rewrite patterns defined in [`SKILL.md`](https://github.com/blader/humanizer/blob/main/SKILL.md) to transform AI-generated text into more natural language.