Skip to main content

Web Chat Interfaces

The simplest way to deploy a security prompt is to paste it directly into the system prompt or custom instructions of a frontier model's web interface. This gives you immediate security guidance with zero infrastructure.

Each setup below starts with the prompt for your technology, for example prompts/python/django/5.2/prompt.md. See Prompt Delivery.

ChatGPT (OpenAI)​

ChatGPT supports persistent custom instructions through Projects, which apply a system prompt to every conversation within that project.

Setup with Projects​

  1. Open chatgpt.com and click Projects in the left sidebar
  2. Create a new project (e.g., "Secure Express Development")
  3. Open the project, select the more options menu (•••), then select Project settings
  4. Paste the full Manicode security prompt into the instructions field
  5. Every conversation within this project will now follow the security prompt

Tips​

  • Create one project per technology (e.g., "Secure Django", "Secure Gin")
  • Project instructions override your account-wide custom instructions inside that project
  • Project instructions persist across sessions and apply to all conversations in that project

Claude (Anthropic)​

Claude supports persistent system prompts through Projects on claude.ai.

Setup with Projects​

  1. Open claude.ai and navigate to Projects
  2. Create a new project (e.g., "Secure Java Development")
  3. Open the project and click Set project instructions
  4. Paste the full Manicode security prompt into the instructions field
  5. All conversations within this project will use the security prompt as context

Tips​

  • Claude's Projects also support uploading files as project knowledge — you can upload the prompt as a .md file alongside relevant code files for additional context
  • Claude has a large context window, so even the longest security prompts will not crowd out your working context
  • You can share projects with team members so everyone uses the same security baseline

Google AI Studio / Gemini​

Google AI Studio provides a System Instructions field that accepts persistent prompts.

Setup​

  1. Open aistudio.google.com and start a new chat in the Playground
  2. Open the System instructions field in the settings panel
  3. Paste the full Manicode security prompt
  4. All messages in this conversation will follow the security rules

Tips​

  • Save your system instructions as a template for reuse
  • Google AI Studio supports structured output — pair the security prompt with specific output formatting instructions if needed
  • For Gemini in Google Workspace, system prompt configuration varies by integration; check your admin console for custom instruction settings

Other Web Interfaces​

Most frontier model web interfaces follow the same pattern. Look for one of these settings:

InterfaceWhere to paste the prompt
PerplexityCustom instructions in a Space
Mistral Le ChatInstructions in a Project
HuggingChatSystem Prompt field in Settings, set per model
OpenRouter ChatSystem prompt field in the chat playground

General Pattern​

  1. Look for a System Prompt, Custom Instructions, or Project Instructions setting
  2. Paste the entire Manicode security prompt
  3. Save or apply the configuration
  4. All subsequent messages will be processed with the security context

Limitations of Web Chat Deployment​

Web chat deployment is the easiest method but has trade-offs:

  • Manual updates — When a prompt update arrives, you must replace the prompt text by hand
  • No version tracking — There is no built-in way to know which prompt version is active
  • Per-user configuration — Each team member must configure their own interface
  • No audit trail — You cannot verify that prompts are being used consistently

For version tracking and shared configuration, commit prompts to your repository instead; see Team and Enterprise Deployment.