prompt-engineering

When to Use

Trigger this skill when designing high-quality prompts for LLMs, optimizing existing prompts, or building prompt template systems.

Core Methods

Prompt Structure

[Role] You are a {domain} expert
[Context] {background information}
[Task] {specific instruction}
[Constraints] {format/length/style requirements}
[Examples] {input → output pairs}
[Output Format] {JSON / Markdown / custom template}

Few-shot Examples

• 2-5 examples optimal, too many causes interference
• Cover edge cases, not just simple scenarios
• Consistent output format is more effective than diverse examples

Chain-of-Thought (CoT)

• Add “Let’s think step by step” or explicitly request step-by-step reasoning
• Break complex tasks into sub-problems, reason sequentially
• Require analysis before conclusion to reduce hallucination

Output Control

• Strict format: use JSON Schema constraints
• Multi-step tasks: ReAct pattern (Reason → Act → Observe)
• Add “If uncertain, say ‘I am not sure’” to reduce fabrication