Skip to content

SKILL.md

Latest commit

 

History

History
103 lines (77 loc) · 2.96 KB

SKILL.md

File metadata and controls

103 lines (77 loc) · 2.96 KB
Raw
name prompting
description Prompt engineering guidance for writing and improving LLM prompts. Use when asked to (1) write a prompt for a specific task, (2) review or improve an existing prompt, (3) design system prompts for AI assistants, (4) structure prompts for specific output formats (JSON, XML, markdown), or (5) apply prompt engineering techniques like few-shot, chain-of-thought, or role prompting.

Prompt Engineering

Guide for crafting effective prompts for large language models (Claude, GPT, Gemini, Llama, etc.).

Workflow

  1. Clarify the goal: What task should the prompt accomplish?
  2. Choose techniques: Select from references/techniques.md
  3. Structure the prompt: Apply appropriate format
  4. Add constraints: Specify requirements and boundaries
  5. Test and refine: Iterate based on outputs

Writing a New Prompt

Start with this template:

[Context/Role - optional]
[Task - required]
[Constraints/Requirements - as needed]
[Output format - as needed]
[Examples - for complex tasks]

Minimal prompt (simple tasks):

Summarize this article in 3 bullet points.

Structured prompt (complex tasks):

You are a senior code reviewer.

Review this code for:
- Security vulnerabilities
- Performance issues
- Maintainability concerns

Format your response as:
## Summary
[1-2 sentences]

## Issues
- [severity]: [description]

## Recommendations
[prioritized list]

Improving an Existing Prompt

Diagnose issues:

Problem Solution
Output too vague Add specific constraints or examples
Wrong format Specify output structure explicitly
Missing details Use chain-of-thought or decomposition
Inconsistent results Add few-shot examples
Off-topic responses Strengthen role/context framing

Improvement checklist:

  • Is the task clear and unambiguous?
  • Are constraints specific (not "be concise" but "under 100 words")?
  • Does output format match intended use?
  • Would examples clarify expectations?

Quick Reference: Techniques

Technique When to Use
Few-shot Specific format/style needed
Chain-of-thought Complex reasoning, math, analysis
Role prompting Domain expertise, specific tone
Task decomposition Multi-step workflows
Constraints Precise requirements

See references/techniques.md for detailed patterns and examples.

Quick Reference: Output Formats

Format When to Use
XML tags Complex prompts, clear section boundaries
JSON Programmatic parsing, structured data
Markdown Human-readable reports, documentation

See references/structured.md for format patterns.

System Prompts

For designing AI assistant behavior, see references/system-prompts.md.

Key sections:

  • Identity and role definition
  • Behavioral guidelines
  • Constraints and boundaries
  • Output format defaults