Prompt Engineering for Agentic Systems

Generate optimized prompts for agentic systems with clear rationale for technique selection.

Workflow

When the user requests a prompt for an agent (e.g., "create a prompt for my email processing agent"):

•Analyze the request - Identify agent type, task complexity, tools available
•Select techniques - Use the decision tree to choose appropriate prompting techniques
•Generate the prompt - Build from the canonical template
•Explain the rationale - Detail why each technique was chosen, including trade-offs

Quick Decision Tree

For fast technique selection:

Agent Characteristic	Recommended Technique
Uses tools autonomously	ReAct
Planning/strategy with alternatives	Tree of Thoughts
High-stakes correctness	Self-Consistency
Factual accuracy, hallucination reduction	Chain-of-Verification (CoVe)
Single-path complex reasoning	Chain of Thought
Complex decisions with trade-offs	Structured Thinking Protocol
Reducing bias, multiple viewpoints	Multi-Perspective Prompting
Uncertainty quantification	Confidence-Weighted Prompting
Proprietary documentation, prevent hallucinations	Context Injection with Boundaries
High-quality content refinement	Iterative Refinement Loop
Strict technical requirements	Constraint-First Prompting
Requires consistent format/tone	Few-Shot (supports negative examples)
Simple, well-defined task	Zero-Shot
Domain-specific expertise	Role Prompting
Procedural workflow	Instruction Tuning

For the full decision tree with detailed branching logic, see decision-tree.md.

Technique Reference

All available techniques with examples, use cases, and risks: techniques.md

Key techniques:

•ReAct - For tool-using agents (reasoning + acting loop)
•Chain of Thought - For complex reasoning tasks
•Chain-of-Verification (CoVe) - For reducing hallucinations and ensuring factual accuracy
•Few-Shot - For enforcing patterns and format (supports negative examples)
•Role Prompting - For defining agent scope and expertise
•Tree of Thoughts - For planning and strategy
•Self-Consistency - For high-stakes verification
•Zero-Shot - For simple, well-defined tasks
•Instruction Tuning - For procedural workflows
•Structured Thinking Protocol - For complex decisions with Understand → Analyze → Strategize → Execute
•Multi-Perspective Prompting - For reducing bias through multiple viewpoints
•Confidence-Weighted Prompting - For uncertainty quantification
•Context Injection with Boundaries - For proprietary documentation to prevent hallucinations
•Iterative Refinement Loop - For high-quality content through multiple passes
•Constraint-First Prompting - For strict technical requirements
•Meta-Prompting - The nuclear option: AI generates its own perfect prompt, then executes it

Anti-Patterns

Common mistakes to avoid: anti-patterns.md

Critical warnings:

•Do NOT use ReAct without tools available
•Do NOT use Tree of Thoughts for deterministic problems
•Do NOT use vague roles ("expert" without scope)
•Do NOT omit stop conditions or error handling

Canonical Template

Use this template as the foundation for all generated prompts: template.md

Rationale Template

When explaining the generated prompt, use this structure:

markdown

## Generated Prompt for [Agent Name/Type]

[prompt in code block]

## Rationale

**Agent Type**: [e.g., Tool-using, Planner, Conversational]

**Task Complexity**: [Simple / Multi-step / Planning-heavy]

**Techniques Used**:
- [Technique]: Why it works for this agent/use case
- [Technique]: Why it works for this agent/use case

**Expected Behavior**: [What the agent will do]

**Trade-offs**: [Cost, latency, flexibility considerations]

**Considerations**: [Edge cases, limitations, or risks]

Guardrail Rule

If a prompt increases latency, token usage, or operational cost, this MUST be stated explicitly in the rationale section under "Trade-offs".