9 AI Skills: Master Prompts, Context, & System Prompts

Prompt engineering is the skill of crafting effective prompts to elicit the best possible output from AI models. A well-structured prompt includes four key components: defining the AI's role (e.g., 'act as a marketer'), providing comprehensive context about your situation, giving clear and specific commands, and specifying the desired output format (e.g., PDF, bullet points, table).

Providing an example of world-class output for the AI to pattern match against acts as a significant 'cheat code' for achieving superior results. This structured approach ensures that the AI understands the user's intent and delivers relevant, high-quality responses, avoiding the 'garbage in, garbage out' scenario.

Taste curation is the ability to discern what constitutes 'great' output from AI, whether it's in visual, auditory, or textual forms. This skill is developed by creating a 'taste library' of exemplary work, such as watching successful startup pitches, reviewing top-tier code on GitHub, or analyzing acclaimed musical performances.

Effective communication skills are paramount, as the words used to interact with AI directly shape its output. Being specific (e.g., 'max line length of 100 characters') and choosing precise vocabulary (e.g., 'leader' instead of 'boss') ensures the AI aligns with your refined taste. Implementing universal rules, such as 'write in ninth-grade English' or 'use similes over examples,' further refines AI responses to consistently meet desired standards.

A master prompt is a comprehensive document that introduces your identity, role, and all relevant context to the AI, ensuring personalized and highly relevant responses. This document acts as a digital ID, uploaded to AI platforms to provide a consistent foundation for all interactions. This approach significantly boosts efficiency, with teams reporting up to 92% of their work supported by AI due to hyper-personalized outputs.

To create a master prompt, instruct the AI to act as an interviewer and ask all necessary questions for your role. Answer these questions in detail, using voice-to-text features to ramble and provide extensive information. Finally, command the AI to generate and save the master prompt as a PDF, which future-proofs it for use across current and future AI platforms like Claude, Gemini, or Grock.

Output iteration involves continuously refining AI-generated content to achieve perfection, rather than settling for 'good enough.' This process requires persistent engagement with the AI, providing specific feedback beyond vague instructions like 'make it punchier.' Instead, specify desired changes, such as 'open with a strong reframe that talks about X, Y, and Z.'

Utilize features like 'canvas' in AI platforms (e.g., ChatGPT) to manually tweak outputs without losing the overall structure. Canvas allows for direct editing of words, sentences, and paragraphs within the AI interface, enabling precise adjustments. This iterative approach, exemplified by Coca-Cola's 70,000 prompts for a single commercial, ensures the final output perfectly matches the user's vision.

While a master prompt defines 'who you are' to the AI, a system prompt instructs the AI 'how to behave.' System prompts are essentially word-based programming, making anyone who can speak a language an AI programmer. To create one, provide the AI with a final, iterated output and ask it to generate the system prompt that would have produced that exact result.

Save this system prompt as a PDF for portability across different AI platforms. For advanced users, this code can be integrated into a custom GPT, transforming it into a reusable component. Custom GPTs, like the 'book architect' example, can perform complex tasks such as extensive research and content generation based on specific inputs, and can be shared with team members for consistent application.

AI can be used as a critical tool to expose blind spots and challenge assumptions, moving beyond its default 'people-pleaser' tendency. By instructing the AI to act as a devil's advocate, it can stress-test ideas and identify risks, providing a harsh reality check that leads to better decision-making. For instance, using it to rebalance an investment portfolio can reveal flaws in personal biases.

To effectively use AI as a critic, ask it to break down its criticism through the lens of first principles, a physics concept that decomposes issues to their fundamental truths. This not only provides critical feedback but also teaches a more rigorous way of thinking. Any valuable insights gained from this critique should be captured and used to update your master prompt, ensuring future AI interactions incorporate these learned principles.

It is possible to provide AI with too much information, leading to confusion and suboptimal outputs. For instance, attempting to train an AI system with 2 million words of context can overwhelm it. The solution is context compression, where AI is used to pre-process and condense vast amounts of information into a manageable size, such as reducing 2 million words to 200,000.

To perform context compression, paste all raw data (transcripts, documents) into the AI and instruct it to summarize key facts, data, and stories, reducing the content to a specific percentage (e.g., 10% of its original size). Subsequently, ask the AI what information was omitted during compression, providing an option to retrieve necessary details. The compressed knowledge then serves as the sole context for subsequent prompts in a new AI session, ensuring focused and efficient processing.

Maintaining an organized AI knowledge base is crucial for consistent and high-quality outputs, akin to weeding a garden. Without proper organization, AI interactions can become messy, with inconsistent data and varied training approaches across teams. Implementing a structured system ensures clarity and reusability.

Create project folders for each initiative, clearly naming them (e.g., 'Investments'). Within each project folder, upload your master prompt and any compressed context relevant to that project. Additionally, keep your best system prompts organized by department in PDF format. This strategy future-proofs your AI workflow, allowing seamless transition between different AI platforms and ensuring consistent application of best practices.

AI serves as a powerful tool for personalized education, enabling users to learn complex topics at their own pace and preferred style. Instead of traditional learning methods, AI can generate custom research papers or summaries on any subject, tailored to specific time constraints and comprehension levels. For example, one can request a 7-minute explanation of AI.

To leverage AI for personalized learning, provide a simple prompt specifying the topic and desired learning duration. Then, instruct the AI to present the information in a simple, conversational language, often by specifying a grade level (e.g., 'seventh-grade language'). The AI can then generate a research paper or summary, which can be listened to using text-to-speech features, transforming daily activities like gym workouts into learning opportunities. AI is not just a chatbot; it's a co-creative operating system for continuous personal and professional development.