# What Is the Default Prompt for the OpenAI Agent in Humanizer? > Discover the default prompt for the OpenAI agent in Humanizer. Learn how to use this prompt to rewrite text in your unique voice while preserving factual accuracy. Find the configuration in openai.yaml. - Repository: [Siqi Chen/humanizer](https://github.com/blader/humanizer) - Tags: how-to-guide - Published: 2026-09-11 --- **The default prompt for the OpenAI agent in Humanizer is "Use $humanizer to rewrite this text in my voice without changing its facts," defined in the [`agents/openai.yaml`](https://github.com/blader/humanizer/blob/main/agents/openai.yaml) configuration file.** The blader/humanizer repository provides a modular framework for humanizing AI-generated content through configurable agents. The OpenAI agent’s default prompt serves as the initial system instruction that directs the language model to preserve factual accuracy while adapting text to match the user’s voice. ## Where the Default Prompt Is Defined The OpenAI agent’s default prompt is stored in **[`agents/openai.yaml`](https://github.com/blader/humanizer/blob/main/agents/openai.yaml)** at the repository root. This YAML file contains the agent metadata that Humanizer loads at runtime. ```yaml default_prompt: "Use $humanizer to rewrite this text in my voice without changing its facts." ``` This configuration file also declares other metadata fields such as `display_name` and `short_description`, but the `default_prompt` key provides the critical instruction set that initializes the agent’s behavior according to the source code. ## How the Default Prompt Works When an OpenAI-compatible agent initializes, the Humanizer loader reads [`agents/openai.yaml`](https://github.com/blader/humanizer/blob/main/agents/openai.yaml) and extracts the `default_prompt` value to use as the system message for the language model. ### Variable Substitution Architecture The default prompt contains a **`$humanizer`** placeholder that the runtime expands dynamically. During execution, Humanizer replaces `$humanizer` with the full skill prompt content stored in **[`SKILL.md`](https://github.com/blader/humanizer/blob/main/SKILL.md)**. This indirection allows multiple agents to reuse the same core humanization logic while maintaining distinct wrapper prompts. The workflow follows this sequence: 1. The loader reads [`agents/openai.yaml`](https://github.com/blader/humanizer/blob/main/agents/openai.yaml) and retrieves the default prompt string. 2. The runtime substitutes `$humanizer` with the content from [`SKILL.md`](https://github.com/blader/humanizer/blob/main/SKILL.md). 3. The expanded prompt becomes the system instruction for the OpenAI API call. Thus, the default prompt acts as a bridge between the agent’s system message and the Humanizer skill logic. ## Customizing and Overriding the Default Prompt While the default prompt functions automatically, you can override it at runtime through CLI arguments or custom configuration files. ### Overriding via CLI When invoking the skill through the command line interface, supply a custom prompt using the **`--prompt`** flag: ```bash npx skills add blader/humanizer --agent openai \ --prompt "Please rewrite this passage in a conversational tone." ``` If you omit the `--prompt` parameter, the agent automatically falls back to the default prompt defined in [`agents/openai.yaml`](https://github.com/blader/humanizer/blob/main/agents/openai.yaml). ### Custom Configuration with Variable Reuse You can also define custom prompts in your own YAML configurations while still leveraging the `$humanizer` variable: ```yaml default_prompt: "Please apply $humanizer to the following text while keeping technical terms unchanged." ``` This approach preserves access to the core humanization engine defined in [`SKILL.md`](https://github.com/blader/humanizer/blob/main/SKILL.md) while tailoring the instruction context for specific use cases. ## Summary - The default prompt for the OpenAI agent in Humanizer is defined in **[`agents/openai.yaml`](https://github.com/blader/humanizer/blob/main/agents/openai.yaml)** as "Use $humanizer to rewrite this text in my voice without changing its facts." - The **`$humanizer`** variable acts as a placeholder that expands to the full skill content from **[`SKILL.md`](https://github.com/blader/humanizer/blob/main/SKILL.md)** at runtime. - You can override the default prompt using the **`--prompt`** CLI flag when invoking the agent. - The prompt architecture enables code reuse across agents while allowing customized instruction contexts. ## Frequently Asked Questions ### Where is the OpenAI agent default prompt stored in the Humanizer repository? The default prompt is stored in **[`agents/openai.yaml`](https://github.com/blader/humanizer/blob/main/agents/openai.yaml)** at the repository root. This file contains the agent’s metadata including the `default_prompt` key that Humanizer reads during initialization to set the system instruction. ### What does the $humanizer variable do in the default prompt? The **`$humanizer`** variable is a placeholder that the Humanizer runtime replaces with the complete skill instructions defined in **[`SKILL.md`](https://github.com/blader/humanizer/blob/main/SKILL.md)**. This substitution happens at execution time, allowing the concise default prompt to access the full humanization logic without duplicating content across agent configurations. ### Can I change the default prompt without modifying the source code? Yes. You can override the default prompt at runtime using the **`--prompt`** flag in the CLI. For example: `npx skills add blader/humanizer --agent openai --prompt "Your custom instructions here"`. If you omit this flag, the agent uses the default prompt from [`agents/openai.yaml`](https://github.com/blader/humanizer/blob/main/agents/openai.yaml). ### How does the default prompt preserve factual accuracy? The default prompt explicitly instructs the model to rewrite text "without changing its facts." This constraint is hardcoded into the prompt string in [`agents/openai.yaml`](https://github.com/blader/humanizer/blob/main/agents/openai.yaml), ensuring the humanization process focuses solely on style and voice adaptation while maintaining the original information integrity.