Everything you need to
build with AI
System prompts, production RAG agents, and app-building megaprompts — organised by what you're trying to create.
System prompts
Model-specific prompts for Claude, Gemini, ChatGPT, Replit and v0. Paste into the system field and start building.
RAG agents
Production-ready agents with full Python source code. Chat with your documents, emails, codebase, and more.
App megaprompts
Fill in the brackets, paste into Claude or Replit — get a working full-stack application.
Prompt engineering frameworks
Reusable structures that work across all models and use cases
RCTF Framework
Best for: Any task — the universal structureRole + Context + Task + Format. The foundation of every good prompt.
🧠Chain of Thought
Best for: Complex reasoning, maths, logic, decisionsForce the model to think step by step before giving an answer.
📋Few-Shot Prompting
Best for: Getting consistent output format, classification, transformation tasksGive 2-3 examples of input → output before your actual request.
📌XML Tags (Claude-specific)
Best for: Claude — complex prompts with multiple sectionsClaude processes XML tags with higher accuracy. Use tags to separate concerns.
✅Self-Consistency Check
Best for: High-stakes decisions, factual claims, code reviewAsk the AI to verify its own output before delivering it.
🚫Negative Prompting
Best for: Writing, code style, formatting controlExplicitly state what NOT to do — often more effective than positive instructions.
🎯RGB — Role, Goal, Format
Best for: Any task needing a specific output typeDefine the AI role, state the goal, specify the output format. Covers 80% of everyday prompting needs in under 30 seconds.
🔬CRISPE Framework
Best for: Complex expert-level analysis and creative tasksCapacity, Role, Insight, Statement, Personality, Experiment — for nuanced expert-level responses.
⭐STAR Prompting
Best for: Real-world debugging, advice, and situational helpStructure your prompt like a behavioural interview answer. Forces specificity about context, need, and desired outcome.
🧅PEEL — Point, Evidence, Explain, Link
Best for: Essay writing, analysis, persuasive contentMakes AI give structured analytical responses: state a point, support with evidence, explain reasoning, link back to question.
👥Expert Panel Prompting
Best for: Important decisions and strategy stress-testingAsk the AI to simulate a panel of experts with different viewpoints. Forces balanced, multi-perspective analysis.
🚧Constraint Prompting
Best for: Fixing generic, off-topic, or bloated outputsTell AI what NOT to do and what hard limits to respect. Constraints fix most quality problems faster than adding more context.
🏛️Socratic Prompting
Best for: Plan stress-testing and complex problem solvingInstead of asking for an answer, ask AI to ask you questions that reveal gaps in your thinking before giving advice.
⚓Output Anchoring
Best for: Consistent formatting across many outputsGive the AI an example of your ideal output before making the request. It mirrors structure, tone, and depth of your anchor.
🔄Iterative Refinement Protocol
Best for: Important emails, documents, product descriptionsThree-pass system: generate, self-critique, refine. The AI critiques its own output then improves it. Dramatically better results.
🎭Persona Injection
Best for: Brand voice consistency and expert-style writingDescribe a persona in detail: beliefs, communication style, vocabulary, what they avoid. AI embodies that persona consistently.
🧩Task Decomposition Prompting
Best for: Long multi-step tasks like building systems or docsAsk AI to break complex tasks into steps, confirm the plan with you, then execute step-by-step. Prevents errors on long tasks.
🪜Prompt Ladder
Best for: When unsure how much context the AI needsA prompting methodology: start 1-sentence prompt, layer in role, context, format, constraints until quality is right.