Prompt Engineering Instructional Language (PEIL)
PEIL is a structured methodology for generating high-quality system prompts for AI agents and large language models.
When to Use This Skill
- •Creating system prompts for autonomous agents
- •Improving existing prompts for better LLM performance
- •Applying research-backed prompting techniques
- •Structuring complex instructions clearly
PEIL Template Structure
Use the following variables to construct effective prompts:
code
{Role} {ProvideClearContext} {BreakDownComplexQuestions} {ProvideSpecificInstructions} {DefineConciseness} {PromptingTechniques} {StateDesiredOutput}
Variable Definitions
| Variable | Purpose |
|---|---|
| Role | Specify the persona or expertise the model should adopt |
| ProvideClearContext | Set the domain and focus area for precise, tailored responses |
| BreakDownComplexQuestions | Decompose complex topics into manageable sub-questions |
| ProvideSpecificInstructions | Define constraints, requirements, or rules |
| DefineConciseness | Set word limits or brevity requirements |
| PromptingTechniques | Apply research-backed techniques (see TECHNIQUES.md) |
| StateDesiredOutput | Specify the expected format, structure, or content |
Hybrid Prompt Structure
Based on research from arXiv:2503.06926, employ a hybrid structure:
- •Opening Statement: Short sentence or paragraph stating role and goal
- •Bullet Points: Specific rules, options, or constraints
Step-by-Step Instructions
- •Define the Role: Start with "You are a [domain] expert" or similar
- •Set Context: Explain the domain and what the model should focus on
- •Break Down Questions: If complex, split into sub-questions
- •Add Constraints: Include any specific rules or requirements
- •Set Length: Define word limits if needed
- •Choose Technique: Select from TECHNIQUES.md based on task type
- •Specify Output: Define format (Markdown, JSON, bullet points, etc.)
Quick Reference: Technique Selection
| Task Type | Recommended Technique |
|---|---|
| Reasoning & Logic | Chain-of-Thought (CoT), Tree-of-Thoughts |
| Reduce Hallucination | RAG, Chain-of-Verification (CoVe) |
| Code Generation | Scratchpad, Program of Thoughts (PoT) |
| New Tasks | Zero-Shot or Few-Shot Prompting |
| Complex Analysis | Decomposed Prompting, Thread of Thought |
See TECHNIQUES.md for the complete techniques table.
Example Prompt
code
You are a cybersecurity expert specializing in enterprise security architecture, focused on practical measures for protecting sensitive data in cloud environments. When responding: - Include at least three key components of a strong security strategy - Provide specific examples of implementation - Address both technical and human factors Think through each security layer step by step. Provide a structured Markdown response with clear headings, limited to 300 words.
Categories
PEIL supports 24 prompt categories. See CATEGORIES.md for full definitions:
Argument, Assessment, Calculation, Categorising, Classification, Clustering, Comparison, Context Control, Contradiction, Cross Boundary, Decomposed Prompting, Error Identification, Hypothesise, Input Semantics, Logical Reasoning, Output Customisation, Output Semantics, Prediction, Prompt Improvement, Refactoring, Requirements Elicitation, Simulation, Summarising, Translation