Prompt Engineering Cheat Sheet 2026 (All Frameworks)

Bookmark this page. Here's every major prompting framework in one place—what they do, when to use them, and copy-paste templates for each.

Quick Reference Table

| Framework | Best For | Complexity | Key Phrase | |-----------|----------|------------|------------| | RISE | Structured content | Low | Role → Instructions → Steps → Expectations | | RACE | Context-heavy tasks | Low | Role → Action → Context → Expectations | | Chain-of-Thought | Reasoning problems | Low | "Let's think step by step" | | Tree-of-Thought | Strategic decisions | Medium | "Explore 3 approaches" | | ReAct | Tasks requiring actions | High | Reason → Act → Observe loop | | Self-Consistency | High-stakes accuracy | Medium | Multiple attempts → consensus | | Least-to-Most | Complex multi-step problems | Medium | Decompose → solve parts → combine |

RISE Framework

Purpose: Structured output with clear role and formatting

Components:

• Role: Who the AI should be
• Instructions: The specific task
• Steps: How to approach it
• Expectations: Output requirements

Template:

Role: You are a [specific expert/professional].

Instructions: [Clear task description].

Steps:
1. [First action]
2. [Second action]
3. [Third action]

Expectations: [Output format, length, style requirements].

Best for: Content creation, expert analysis, formatted outputs

RACE Framework

Purpose: Context-rich tasks requiring situational awareness

Components:

• Role: Who the AI should be
• Action: What to do
• Context: Background and constraints
• Expectations: Desired outcome

Template:

Role: You are a [specific role].

Action: [What you need done].

Context:
- [Key background point 1]
- [Key background point 2]
- [Constraints or special considerations]

Expectations: [What success looks like].

Best for: Business communications, personalized content, nuanced situations

Chain-of-Thought (CoT)

Purpose: Improve reasoning accuracy by showing work

Basic Template:

[Problem statement]

Let's solve this step by step.

Detailed Template:

[Problem statement]

Work through this systematically:
1. First, identify what we know
2. Then, determine what we need to find
3. Apply the relevant method
4. Verify the answer makes sense

Show your reasoning at each step.

Best for: Math, logic, debugging, analysis, any task where accuracy matters

Tree-of-Thought (ToT)

Purpose: Explore multiple solutions before committing

Template:

[Problem or decision]

Generate 3 different approaches to this:

For each approach:
1. Describe the strategy
2. List pros and cons
3. Identify key risks
4. Rate confidence (1-10)

Then recommend the best approach or propose a hybrid solution.

Best for: Strategic decisions, creative problems, complex trade-offs

ReAct (Reason + Act)

Purpose: Tasks requiring external actions or information gathering

Template:

Task: [Description]

Available tools:
- [tool1](params): [Description]
- [tool2](params): [Description]

Use the ReAct framework:
1. Thought: What do I need to do/know?
2. Action: [tool_name](parameters)
3. Observation: [wait for result]
4. Repeat until complete

Final Answer: [Conclusion]

Best for: Research, fact-checking, multi-step workflows, AI agents

Self-Consistency

Purpose: Increase accuracy through multiple attempts

Template:

[Problem statement]

Solve this problem 3 times using different approaches. Then compare your answers and provide the most likely correct answer with your confidence level.

Alternative Template:

[Question]

Approach 1: [Solve using method A]
Approach 2: [Solve using method B]
Approach 3: [Solve using method C]

Consensus: Compare results and determine the most reliable answer.

Best for: High-stakes decisions, complex calculations, when accuracy is critical

Least-to-Most

Purpose: Break complex problems into manageable subproblems

Template:

[Complex problem]

First, break this into smaller subproblems:
1. [Subproblem 1]
2. [Subproblem 2]
3. [Subproblem 3]

Now solve each subproblem in order, using previous solutions to inform the next.

Finally, combine the solutions to solve the original problem.

Best for: Multi-step problems, complex projects, learning progressions

Framework Decision Tree

START
  │
  ├─ Need structured content/output?
  │     ├─ Yes, with heavy context → RACE
  │     └─ Yes, with clear steps → RISE
  │
  ├─ Need to reason through a problem?
  │     ├─ Single path solution → Chain-of-Thought
  │     ├─ Multiple options to explore → Tree-of-Thought
  │     └─ Accuracy is critical → Self-Consistency
  │
  ├─ Need to take actions/gather info?
  │     └─ Yes → ReAct
  │
  └─ Problem is very complex?
        └─ Yes → Least-to-Most

Quick Combinations

Strategy + Execution

Phase 1 (ToT): "Generate 3 approaches to [goal]. Evaluate and select best."
Phase 2 (RISE): "Now implement the chosen approach with this structure..."

Research + Analysis

Phase 1 (ReAct): Gather data using available tools
Phase 2 (CoT): "Analyze this data step by step..."

Context + Reasoning

RACE setup + "Now work through this step by step" (CoT)

Token Cost Comparison

| Framework | Relative Cost | When Worth It | |-----------|--------------|---------------| | Basic prompt | 1x | Simple tasks | | RISE/RACE | 1.2x | Most content tasks | | Chain-of-Thought | 1.5x | Accuracy-sensitive tasks | | Tree-of-Thought | 3-4x | Important decisions | | Self-Consistency | 3x | High-stakes accuracy | | ReAct | Variable | Action-based tasks |

Common Mistakes

| Framework | Mistake | Fix | |-----------|---------|-----| | RISE | Vague role | Be specific: "senior fintech product manager" not "business expert" | | RACE | Thin context | Add 3-5 bullet points minimum | | CoT | Using for simple tasks | Skip CoT for basic requests | | ToT | Only 2 branches | Always explore 3+ options | | ReAct | Undefined tools | Specify exact tool names and parameters |

Print-Friendly Summary

RISE = Role + Instructions + Steps + Expectations → Structured content

RACE = Role + Action + Context + Expectations → Context-rich tasks

CoT = "Step by step" → Reasoning problems

ToT = "3 approaches, evaluate, select" → Decisions

ReAct = Think → Act → Observe → Repeat → Tasks requiring actions

Self-Consistency = Multiple attempts → consensus → High accuracy needs

Least-to-Most = Break down → Solve parts → Combine → Complex problems

Want these frameworks applied automatically? PromptWizz selects and applies the optimal framework for any prompt.