Instruction file imported from JohnMichaelMiller/zeus.academia.3b (.github/instructions/prompt-file-generation.instructions.md). Copyright stays with the author.

Prompt File Generation

Naming & Location

<purpose>.prompt.md in .github/prompts/ or context directory

Metadata Requirements

AI Provenance (AI-generated files only)

See ai-assisted-output.instructions.md for required fields (🔒 canonical source).

Prompt-specific additional fields: description, context, expected_output

Copilot Prompt Metadata

Required: name (unique ID) Recommended: description, author, tags: [...] Optional: arguments, examples, context, expected_output, tools: [...], mode

Note: author refers to the creator of the prompt template. operator (in AI provenance) refers to the user executing the AI session.

name: generate-readme
description: Generate comprehensive README
author: John Miller
tags: [documentation, markdown]
arguments:
  - name: project_name
    description: Project name
  - name: language
    description: Programming language
examples:
  - input: "Generate README for Python app"
    output: "README with setup, usage, contribution sections"
context: "Python dev environment"
expected_output: "Markdown README"
tools: ["generate", "edit"]
mode: interactive

Content Structure

1. Title - Clear task
2. Context - Background, constraints, dependencies
3. Instructions - Direct objective, success criteria, format, deliverables
4. Examples - Input/output pairs, edge cases
5. Validation - Quality checklist

Argument Syntax: Use {{variable_name}} in body; define in arguments array

Writing Rules

Specificity: Measurable criteria, explicit actions, defined terms Clarity: Numbered steps (sequences), bullets (items), headings (concerns) Reusability: Placeholders, documented substitutions, separated content Precision: Examples for patterns, explicit format, edge case handling

Patterns

Code Gen:

# Generate <Type>

## Requirements

- Language: <lang>
- Framework: <fw>
- Functions: <list>

## Output Format

<structure>
## Constraints
<limits>

Analysis:

# Analyze <Subject>

## Input

<format>
## Focus
1. <aspect>
## Output
<findings format>

Transform:

# Transform <A> to <B>

## Source/Target

<formats>
## Rules
- <mapping>

Anti-Patterns

❌ Vague goals | ❌ Implicit context | ❌ Missing format | ❌ Over-complexity | ❌ No examples | ❌ Buried requirements

Checklist

• Metadata: AI provenance (if generated) + Copilot name + description/author/tags
• Arguments: Defined in frontmatter, used as {{variable}} in body
• Objective: Clear in first 2 sentences
• Success criteria: Measurable
• Format: Explicit or obvious
• Examples: Provided for complexity
• Edge cases: Addressed
• Testable: As written

Token Target: 300-800 typical | 1500 max complex

Optimization: Front-load critical info | Structure > prose | No filler ("please", "try to") | Examples > descriptions