Powerful Prompts

Battle-tested prompts for sharper AI reasoning, writing, estimation, and adaptive systems—ready to copy, adjust.

Directory Structure

prompts/
├── reasoning/              # Structured reasoning and problem-solving
├── estimation/             # Mathematical and Fermi estimation
├── writing/                # Content creation and review
├── advanced-techniques/    # Complex multi-strategy prompts
└── meta-utility/          # Prompt improvement and adaptive systems

Prompt Map

Reasoning (prompts/reasoning/)

• cot_prompt.txt, reflection_prompt.txt, deepdive.txt, sophisticated_verbal_problem_solving.txt

Estimation (prompts/estimation/)

• fermi.txt

Writing (prompts/writing/)

• summarisation.txt, factual-consistency-summarisation.txt, editorial_review_prompt.txt, generate_expert_qna_prompt.txt, generate-verbal-feedback-in-pairwise-ranking.txt

Advanced Techniques (prompts/advanced-techniques/)

• graph-of-thought-prompting.txt, synthetic_reasoning_prompt.txt, mystery_solver.txt, abstract-prompt.txt, atomic_elements_prompt.txt, prompt_chain_of_symbols.txt

Meta & Utility (prompts/meta-utility/)

• adaptive_dynamic_AI_assistant.txt, prompt_improver_context_meta_codes.md

Quick Start

Basic Usage

1. Pick a prompt from the map above.
2. Copy the file content.
3. Swap placeholder variables (usually {{DOUBLE_BRACES}}).
4. Paste into your AI interface or code.

Example: Using Chain of Thought

# For programmatic use
cot_prompt = """
You are an AI assistant that uses a Chain of Thought (CoT) approach with reflection...
[Full prompt content]
"""

response = model.generate(cot_prompt + "\n\nQuery: " + user_question)

Example: Fermi Estimation

Replace {{FERMI_QUESTION}} with your estimation problem, e.g.
"How many piano tuners are there in Leeds?"

Best Practices

Choosing the Right Prompt

1. Logical reasoning: cot_prompt.txt or reflection_prompt.txt
2. Calculations: fermi.txt or deepdive.txt
3. Writing: any file in prompts/writing/
4. Complex multi-faceted problems: abstract-prompt.txt or graph-of-thought-prompting.txt
5. Adaptive behaviour: adaptive_dynamic_AI_assistant.txt

Optimisation Tips

• Combine prompts (e.g., CoT + Reflection)
• Adjust detail level to match complexity
• Iterate with reflection to refine outputs
• Customise placeholders to your domain

Advanced Usage

Chaining Prompts

1. Break the task with `atomic_elements_prompt.txt`
2. Solve each element with `cot_prompt.txt`
3. Validate with `reflection_prompt.txt`

Prompting Techniques to Try (beyond those in the files)

• Self-consistency sampling: draft several reasoning paths and keep the consensus answer.
• Deliberate planning: outline steps before generating the solution.
• Socratic questioning: ask clarifying questions before answering.
• Few-shot scaffolding: show a mini example for each subtask, not just the final output.
• Program-Aided Logic (PAL): write executable pseudo-code and then explain the result.
• Retrieval-augmented prompting: fetch context (or craft hypotheticals like HyDE) before reasoning.
• Constraint-first prompting: declare hard limits, non-goals, and style rules upfront.
• Persona stacking: rotate through expert roles to stress-test answers.
• Counterfactual stress tests: ask “what would make this fail?” then patch the answer.
• Style transfer prompting: match tone/voice from a short reference excerpt.

Use Cases

Research & Analysis

• Literature review synthesis
• Complex problem decomposition
• Multi-factor decision analysis

Education

• Worked example generation
• Step-by-step problem solving
• Conceptual understanding assessment

Content Creation

• Technical writing
• Article editing and review
• Summary generation

Software Development

• Debugging complex issues
• System design analysis
• Code review and optimisation

Contributing

Contributions are welcome! When adding new prompts:

1. Place new prompts in the right category folder.
2. Use descriptive filenames (lowercase with underscores or hyphens).
3. Explain placeholder variables.
4. Add examples where helpful.
5. Update this README with a short note.

File Naming Conventions

• Use lowercase with underscores or hyphens.
• Be descriptive: chain_of_thought.txt not cot.txt.
• Indicate technique: *_prompt.txt, *_framework.txt.
• Match content to the filename.

License

MIT License - See LICENSE for details

Citation

If you use these prompts in your research or projects, please cite:

@misc{powerful-prompts-2024,
  author = {Roberts, Gareth},
  title = {Powerful Prompts: A Collection of Advanced AI Prompting Techniques},
  year = {2024},
  publisher = {GitHub},
  url = {https://github.com/dr-gareth-roberts/Powerful-Prompts}
}

Additional Resources

• Prompt Engineering Guide
• Chain-of-Thought Paper
• Tree-of-Thought Paper
• Graph-of-Thought Paper

Support

For questions, issues, or suggestions:

• Open an issue on GitHub
• Contribute improvements via pull requests

---

Last Updated: February 2026