# Context Engineering for Non-Developers How to practice [[Context Engineering]] without writing code. For people who use ChatGPT, Claude, or Gemini through the web interface. The principles are the same; the implementation is different. **Custom instructions as your master prompt.** Every major AI platform lets you set persistent instructions. This is your [[AI Master Prompt]] without a codebase. Write who you are, what you do, how you want responses formatted, and what the AI should always know about you. **Projects as context containers.** Claude Projects, ChatGPT Custom GPTs, and similar features let you attach documents as persistent context. This is [[Context Layering]] in practice. Your project description is the system layer; uploaded files are the reference layer; your conversation is the working layer. **Saved conversation starters.** Pre-written prompts for recurring tasks. Instead of re-explaining what you need each time, you load a saved starter that sets up the context. **Uploaded documents as context.** PDFs, markdown files, spreadsheets. These fill the [[Context Budget]] the same way programmatic context injection does. The constraint is the same: you cannot load everything, so you must curate. **The principles transfer directly:** - [[Context Budget]]: you have a limited window; choose what goes in it carefully - [[Context Layering]]: organize context from stable (identity, preferences) to volatile (current task) - [[Context Hygiene]]: stale uploaded documents produce stale output; review and refresh them - [[Levels of AI Context Management]]: you can progress from ad-hoc to systematic without touching code [[AI Assistants]] are the tools. [[Personal Context Management (PCM)]] is the practice. No code required. ## References ## Related - [[Context Engineering]] - [[AI Master Prompt]] - [[Context Budget]] - [[Context Layering]] - [[Context Hygiene]] - [[AI Assistants]] - [[Levels of AI Context Management]] - [[Personal Context Management (PCM)]]