Discover 177+ Techniques & Frameworks

Browse every prompt engineering technique and framework in the Praxis Library, organized by category. New to prompting? Start with Prompt Basics.

Structured Frameworks 5

Step-by-step templates that give every prompt a clear, repeatable structure.

Five-element template: Context, Role, Instructions, Specifics, Parameters.

Still active framework

CRISP extended with Examples for consistent, repeatable outputs.

Still active framework

Six-element template optimized for audience-aware professional content.

Still active framework

Systematic approach to organizing prompt context and structural elements.

Still active technique

Techniques for controlling AI output format, length, and structure.

Still active technique

View Category Page →

Reasoning & CoT 27

Chain-of-thought methods that guide AI through transparent, step-by-step logical reasoning.

Encourage step-by-step reasoning through intermediate thinking steps.

Still active technique

Add "Let's think step by step" without examples for reasoning tasks.

Still active technique

Automatically generate diverse reasoning demonstrations with clustering.

Still active technique

Provide both correct and incorrect reasoning examples for contrast.

Still active technique

Triple-phase reasoning: analysis, structured thinking, summarization.

Still active technique

Symbolic reasoning combined with natural language for verifiable answers.

Still active technique

Select the most complex reasoning paths for harder problems.

Still active technique

Tabular format to organize step-by-step reasoning in columns.

Still active technique

Solve problems by generating and applying relevant analogies.

Still active technique

Abstract to higher-level concepts before solving specific problems.

Still active technique

Segment and reason through lengthy context systematically.

Still active technique

Identify uncertain examples and annotate them for better demonstrations.

Still active technique

Pre-think across unlabeled data before tackling target problems.

Still active technique

Work backward from conclusions to verify reasoning validity.

Still active technique

Apply chain-of-thought selectively based on model confidence for cost-optimal reasoning.

Still active technique

Embed algorithmic search patterns like DFS and BFS directly into prompts for single-call solution-space exploration.

Still active technique

Maintain reusable thought-templates from past problem-solving for template-guided reasoning.

Still active technique

Build recursive explanation trees and use logical consistency to verify commonsense reasoning.

Still active technique

Show intermediate computation steps in a dedicated workspace before producing final answers.

Still active technique

Alternate between selecting relevant premises and drawing conclusions for structured logical reasoning.

Still active technique

Solve analogous problems first and propagate insights to refine the target problem’s solution.

Still active technique

Replace natural language with compact symbolic representations for spatial and structural reasoning.

Still active technique

Integrate formal logic rules and symbolic expressions with chain-of-thought for verifiable reasoning.

Still active technique

Route between fast intuitive responses and slow deliberate reasoning based on task complexity.

Still active technique

Generate structured evidence triples with explanation hints for transparent, verifiable reasoning.

Still active technique

Reason with abstract placeholders first, then fill in concrete values using domain tools.

Still active technique

Combine reinforcement learning and Monte Carlo Tree Search for flexible, efficient reasoning.

Still active technique

View Category Page →

Decomposition 11

Break complex problems into smaller, manageable parts with tree, graph, and recursive approaches.

Break complex problems into progressively harder sub-problems.

Still active technique

Decompose questions and delegate to specialized sub-question handlers.

Still active technique

Generate a plan before executing each step sequentially.

Still active technique

Explore multiple reasoning branches and evaluate alternative paths.

Still active framework

Model reasoning as a graph with merging and refining operations.

Still active framework

Recursively decompose and solve nested sub-problems.

Still active technique

Generate executable code as intermediate reasoning steps.

Still active technique

Generate an outline first, then expand each point in parallel for faster structured responses.

Still active framework

Decompose complex tasks into parallel branches, solve each independently, then merge results.

Still active technique

Decompose complex questions into sequential sub-questions, answering each before proceeding.

Still active technique

Reason over tabular data through step-by-step table operations for verifiable data analysis.

Still active technique

View Category Page →

Self-Correction 11

Techniques that let AI critique, verify, and iteratively refine its own outputs.

Iteratively generate, critique, and refine outputs without external input.

Still active technique

Generate multiple solutions and verify each against the problem.

Still active technique

Plan, execute, and verify a series of fact-checking questions.

Still active technique

Use external tools to verify and correct AI-generated content.

Still active technique

Build answers incrementally through iterative propositions.

Still active technique

Assess confidence in answers to determine reliability.

Still active technique

Reflect on past failures through verbal self-feedback for improvement.

Still active framework

Bootstrap reasoning ability by iteratively generating and fine-tuning on successful reasoning chains.

Still active technique

Train models to generate internal reasoning at every token for improved prediction without explicit prompts.

Still active technique

Verify each reasoning step for errors, then edit the chain to fix mistakes before the final answer.

Still active technique

Use the model’s previous answers as progressive hints to iteratively refine toward the correct solution.

Still active technique

View Category Page →

In-Context Learning 12

Few-shot, zero-shot, and example-selection methods that teach AI by demonstration.

Provide examples to guide AI output format and reasoning patterns.

Still active technique

Direct task instruction without examples, relying on model knowledge.

Still active technique

Single example to establish output pattern and expectations.

Still active technique

Choose the most informative demonstrations for the task at hand.

Still active technique

Select examples nearest to the query in embedding space.

Still active technique

Iteratively select unlabeled examples via model confidence voting.

Still active technique

Combine results from multiple demonstration sets for robustness.

Still active technique

Discover optimal prompt templates through systematic search.

Still active technique

Provide hundreds to thousands of demonstrations using extended context windows.

Still active technique

Optimize the sequence of few-shot examples to maximize model performance.

Still active technique

Have the model generate its own demonstrations before tackling the task.

Still active technique

Dynamically choose demonstrations based on model uncertainty per query.

Still active technique

View Category Page →

Ensemble Methods 12

Combine multiple reasoning paths, prompts, or outputs to achieve more reliable and accurate results.

Sample multiple reasoning paths and take majority vote for accuracy.

Still active technique

Self-generated demonstrations replace human-written examples.

Still active technique

Combine reasoning across multiple prompt perspectives.

Still active technique

Select prompts that maximize information between input and output.

Still active technique

Evaluate and select the best reasoning process for each problem.

Still active technique

Apply self-consistency to free-form text generation tasks.

Still active technique

Diverse verifier on reasoning steps for reliable ensemble answers.

Still active technique

Assign multiple expert personas to analyze a problem from diverse specialized perspectives.

Still active technique

Generate answers from multiple simulated experts with structured discussion and voting.

Still active technique

Have agents argue opposing positions to extract the most well-supported answer.

Still active technique

Multiple reasoning agents share intermediate thoughts for collaborative problem-solving.

Still active technique

Compare outputs in pairs for more consistent and reliable evaluation than absolute scoring.

Still active technique

View Category Page →

Prompting Strategies 33

Versatile techniques spanning agentic workflows, communication styles, and input refinement methods.

Interleave reasoning traces with task-specific actions.

Still active framework

Decompose questions by asking and answering follow-up questions.

Still active technique

Connect multiple focused prompts in sequential workflows.

Still active technique

Assign an expert persona to access specialized knowledge.

Still active technique

Add emotional context to enhance AI response quality and engagement.

Still active technique

Control writing style, tone, and voice characteristics.

Still active technique

Let AI ask clarifying questions before responding.

Still active technique

Regenerate context to remove irrelevant information before answering.

Still active technique

Simulate different perspectives for theory-of-mind reasoning.

Still active technique

Rephrase the question before answering to improve comprehension.

Still active technique

Re-read the question to enhance understanding of complex problems.

Still active technique

Define AI behavior, personality, and constraints through persistent system-level instructions.

Still active technique

Ground AI responses in retrieved documents to reduce hallucination and deliver source-backed answers.

Still active framework

Design AI systems that autonomously plan, execute, and adapt tasks using tools and reasoning loops.

Still active technique

Program language models with typed signatures and automatic compilation instead of hand-crafted prompts.

Still active framework

Optimize multi-stage LM programs with trace-grounded instruction proposals and Bayesian search.

Still active framework

Optimize multi-turn agentic systems in-the-flow with group refined policy optimization.

Still active framework

High-level structural templates that teach HOW to approach tasks rather than providing specific content.

Still active technique

Process multiple task instances in a single prompt for efficiency and cost reduction.

Still active technique

Simply repeating key instructions improves non-reasoning LLM performance through emphasis.

Still active technique

Use targeted hints and cues to steer LLM outputs toward desired directions.

Still active technique

Model generates relevant knowledge before answering to improve accuracy on reasoning tasks.

Still active technique

Generate hypothetical answer documents to improve retrieval without task-specific training.

Still active technique

Transform task inputs into question-answer format and aggregate multiple reformulated responses.

Still active technique

Use reflective Socratic questioning to guide AI through iterative reasoning and discovery.

Still active technique

Use multi-turn dialogue structure to guide reasoning through conversational exchange.

Still active technique

Infer task instructions from a few input-output demonstrations automatically.

Still active technique

Automatically generate and select optimal prompt instructions via LLM-driven search.

Still active technique

Use LLMs as optimizers where the optimization task is described in natural language.

Still active technique

Generate multiple semantically similar prompt variants and ensemble their responses.

Still active technique

Treat reasoning as planning using the LLM as both world model and agent with MCTS.

Still active technique

Synergize reasoning, acting, and planning via Monte Carlo Tree Search for agents.

Still active technique

Iteratively refine summaries to be progressively more entity-dense without increasing length.

Still active technique

View Category Page →

Code 11

Specialized techniques for generating, debugging, and structuring code outputs.

Techniques for generating, explaining, and transforming code effectively.

Still active technique

Let AI debug its own generated code through explanation and testing.

Still active technique

Generate formatted outputs like JSON, XML, and structured tables.

Still active technique

Guide AI to generate complete, functional programs from high-level specifications and input-output examples.

Still active technique

Prompt AI to analyze, interpret, and explain existing code at multiple levels of abstraction.

Still active technique

Direct AI models to perform systematic code reviews covering security, performance, and maintainability.

Still active technique

Guide AI to create comprehensive test suites with unit tests, integration tests, and edge case coverage.

Still active technique

Translate natural language questions into accurate, optimized SQL queries across database systems.

Still active technique

Interweave code execution with LM simulation for reasoning across executable and semantic tasks.

Still active technique

Generate Python programs as reasoning steps and offload execution to an interpreter.

Still active technique

SQL-like query language for structured, constrained interaction with language models.

Still active technique

View Modality Hub →

Safety & Alignment 3

Methods for aligning AI behavior, preventing misuse, and ensuring safe deployment.

Use a set of principles to guide AI self-critique and revision without human feedback.

Still active method

Simpler alternative to RLHF that directly optimizes model policy from preference data.

Still active method

Define priority levels for conflicting instructions to prevent prompt injection.

Still active method

Community Frameworks 21

Community-developed structured prompting frameworks for organizing and improving AI interactions.

Four-component framework: Role, Action, Context, Expected Output.

Still active framework

Five-component framework with step sequencing and scope narrowing.

Still active framework

Four-part framework emphasizing clear data input and output specification.

Still active framework

Six-component framework with persona, examples, and iterative adjustments.

Still active framework

Five-element framework ensuring clear objectives with illustrative examples.

Still active framework

Narrative-driven framework for creative problem-solving with a serendipity element.

Still active framework

Five-component framework with task clarity and contextual grounding.

Still active framework

Context-aware framework from Nielsen Norman Group with role assignment and outcomes.

Still active framework

Five-part framework with built-in output validation through sense checking.

Still active framework

Four-component framework emphasizing background context and measurable expectations.

Still active framework

Beginner-friendly three-part framework: Action, Purpose, Expectation.

Still active framework

Three-element framework prioritizing expected outcomes before role and action.

Still active framework

Minimalist three-part framework for persona, task, and output format.

Still active framework

Goal-oriented three-element framework clarifying what, how, and why.

Still active framework

Narrative framework: Before (current state), After (desired state), Bridge (solution path).

Still active framework

Five-component framework combining OKR methodology with prompt engineering.

Still active framework

Five-component framework emphasizing scenario context and step-by-step guidance.

Still active framework

Educational framework integrating chain-of-thought into structured prompting.

Still active framework

Goal-setting framework adapted for AI prompting from project management methodology.

Still active framework

Self-optimizing agent prompt evolution framework for autonomous refinement.

Still active framework

Living document meta-framework providing comprehensive organizational context to AI.

Still active framework

Image Prompting 12

Techniques for understanding, generating, editing, and composing images with AI models.

Foundational techniques for prompting AI to understand, analyze, and reason about images.

Still active technique

Extend chain-of-thought reasoning to combine visual and textual information step by step.

Still active technique

Guide models to reason through images region by region in explicit visual steps.

Still active technique

Convert visual information into structured text descriptions for text-only model reasoning.

Still active technique

Ask precise questions about images and get accurate, grounded answers from multimodal AI.

Still active technique

Craft text prompts that guide AI image generators to produce specific, high-quality visuals.

Still active technique

Specify what to exclude from AI-generated images for higher quality and artifact avoidance.

Still active technique

Use structural conditioning inputs like edge maps and pose skeletons to control generation.

Still active technique

Selectively edit, replace, or restore specific regions within existing images.

Still active technique

Apply artistic styles and visual aesthetics to images through prompt-based techniques.

Still active technique

Transform existing images using reference inputs combined with text prompts.

Still active technique

Control spatial arrangement, framing, and visual hierarchy in AI-generated images.

Still active technique

View Modality Hub →

Audio Prompting 6

Techniques for speech recognition, synthesis, classification, and music generation with AI models.

Foundational techniques for prompting AI to process, understand, and generate audio content.

Still active technique

Guide AI transcription models with context, vocabulary hints, and formatting instructions.

Still active technique

Control voice synthesis with tone, pacing, emotion, and pronunciation directives.

Still active technique

Prompt AI to categorize, tag, and identify patterns in audio signals and recordings.

Still active technique

Craft prompts that guide AI music generators with genre, mood, tempo, and instrumentation.

Still active technique

Techniques for replicating vocal characteristics and creating consistent synthetic voices.

Still active technique

View Modality Hub →

Video Prompting 6

Techniques for video understanding, generation, temporal reasoning, captioning, and editing with AI models.

Foundational techniques for prompting AI to analyze, understand, and reason about video content.

Still active technique

Craft prompts that guide AI to create and synthesize video content from textual descriptions.

Still active technique

Understand and reason about events across time, including causal relationships and ordering.

Still active technique

Ask and answer specific questions about video content using multimodal AI models.

Still active technique

Generate accurate, descriptive captions and text descriptions for video content.

Still active technique

Use natural language to modify, transform, and enhance existing video content with AI.

Still active technique

View Modality Hub →

3D / Spatial 5

Techniques for 3D model generation, scene understanding, pose estimation, and spatial data analysis with AI.

Foundational techniques for guiding AI to understand, generate, and reason about three-dimensional spatial data.

Still active technique

Prompt AI to create three-dimensional models, meshes, and objects from natural language descriptions.

Still active technique

Analyze complete 3D scenes including spatial relationships, object placement, and environmental context.

Still active technique

Guide AI to detect, analyze, and reason about human body poses and skeletal configurations.

Still active technique

Process and analyze 3D point cloud data from LiDAR, depth sensors, and photogrammetry scans.

Still active technique

View Modality Hub →

Which Technique Should I Use?

Answer 3 questions to find your best match, or compare frameworks side by side

Answer 3 questions to find your best match

1/3

What's your primary goal?

Technique	Best For	Key Strength
CRISP	Quick tasks, daily use	Simplicity & speed
CRISPE	Consistent outputs	Few-shot learning
COSTAR	Professional content	Audience awareness
Chain-of-Thought	Complex reasoning	Transparent logic
ReAct	Multi-step reasoning	Action + thought
Self-Consistency	Reliable answers	Majority voting
Tree of Thought	Exploration tasks	Branch evaluation

Full Technique Finder →

New to Prompting?

Start with the Basics

Before diving into frameworks, master the four elements that make every prompt effective: clear instructions, relevant context, input data, and output format.

Prompt Basics Try Prompt Analyzer

Discover 177+ Techniques & Frameworks

Structured Frameworks 5

CRISP

CRISPE

COSTAR

Context & Structure

Constrained Output

Reasoning & CoT 27

Chain-of-Thought

Zero-Shot CoT

Auto-CoT

Contrastive CoT

Structured CoT

Faithful CoT

Complexity-Based

Tab-CoT

Analogical Reasoning

Step-Back Prompting

Thread of Thought

Active Prompting

Memory of Thought

Reversing CoT

Uncertainty-Routed CoT

Algorithm of Thoughts

Buffer of Thoughts

Maieutic Prompting

Scratchpad Prompting

Selection-Inference

Thought Propagation

Chain-of-Symbol

Symbolic CoT

Dual Process Prompting

Chain of Knowledge

Chain of Abstraction

Everything of Thoughts

Decomposition 11

Least-to-Most

DECOMP

Plan-and-Solve

Tree of Thought

Graph of Thought

Recursion of Thought

Program of Thought

Skeleton-of-Thought

Branch-Solve-Merge

Successive Prompting

Chain of Table

Self-Correction 11

Self-Refine

Self-Verification

Chain-of-Verification

CRITIC

Cumulative Reasoning

Self-Calibration

Reflexion

STaR

Quiet-STaR

Verify-and-Edit

Progressive-Hint

In-Context Learning 12

Few-Shot Learning

Zero-Shot

One-Shot

Example Selection

KNN Prompting

Vote-k

Demo Ensembling

Prompt Mining

Many-Shot

Example Ordering

Self-Generated ICL

Active Example Selection

Ensemble Methods 12

Self-Consistency

COSP

Dense Prompting

Max Mutual Information

Meta-Reasoning

Universal Self-Consistency

DiVeRSe