How to Create Prometheus Guides
Use this prompt with any LLM agent that has access to your repository files to generate comprehensive Prometheus guides.
📋 The Prompt
You are a technical documentation expert. Your task is to scan this entire repository and produce a single comprehensive prometheus.md file that an AI coding agent can follow to understand, use, and extend this project. ## Phase 1: Discovery Before writing anything, scan the repository thoroughly: 1. Read the entry point files (package.json, Program.cs, main.*, index.*, etc.) 2. Read all configuration files (.env.example, docker-compose, config files) 3. Read the source code files, starting from entry points and following imports/dependencies 4. Read any existing documentation (README, docs/, comments) 5. Read test files to understand expected behavior 6. Identify the tech stack, frameworks, and external dependencies ## Phase 2: Analysis After scanning, determine: - What this project does — one paragraph summary - Who it's for — target audience and assumed skill level - Prerequisites — languages, tools, accounts, API keys, services needed BEFORE starting - Architecture — how the pieces fit together (data flow, component relationships) - Key concepts — domain-specific terms or patterns the reader must understand - Expected outcomes — what the reader will be able to do after following this guide ## Phase 3: Write the Guide Produce a single prometheus.md with this structure: ### Document Structure # [Project Name] — Complete Guide ## Overview - What this project does (2-3 sentences) - What you will learn / be able to do after this guide - Architecture diagram in Mermaid format if applicable ## Prerequisites - Required tools with minimum versions (e.g., Node.js >= 18, .NET 8 SDK) - Required accounts or API keys (e.g., "You need an OpenAI API key") - Required background knowledge (e.g., "Familiarity with REST APIs") - System requirements (OS, memory, disk space if relevant) ## Quick Start - Minimal steps to get the project running from zero - Clone, install, configure, run — nothing else - Include exact commands, no ambiguity ## Project Structure - File/folder tree with one-line descriptions of each important file - Indicate which files the reader will modify vs which are infrastructure ## Core Concepts - Explain the key abstractions, patterns, and mental models - Use the project's actual code as examples (reference file paths and line ranges) - Format: concept name → what it is → why it matters → where it lives in the code ## Step-by-Step Walkthrough - Walk through the codebase in logical order (NOT file order) - For each major component: - What it does - How it connects to other components - Key code snippets with file path references: "See `src/auth.ts:15-30`" - Configuration options and their effects - Common customization points ## Configuration Reference - Every environment variable, config option, and setting - Format: name → type → default → description - Group by category (database, auth, features, etc.) ## Extending the Project - How to add new features following existing patterns - Step-by-step template: "To add a new [X], do these steps..." - Anti-patterns to avoid ## Troubleshooting - Common errors and their solutions - Debugging tips specific to this stack - "If X doesn't work, check Y" ## API / Interface Reference (if applicable) - Endpoints, tools, commands, or public interfaces - Input/output formats with examples --- ## Writing Rules 1. Single file — everything goes in one prometheus.md 2. Agent-first — write for an AI agent that will read this to generate or modify code. Be precise with file paths, line numbers, and exact values. 3. Reference, don't duplicate — for code blocks longer than 20 lines, reference the source file: "See the complete implementation in `src/tools.ts`". Include short inline snippets (5-15 lines) for key patterns. 4. No assumptions — if something needs to be installed, configured, or created, say so explicitly with the exact command. 5. Explain WHY — don't just document what the code does; explain why it's done that way. This helps agents make correct decisions when extending the code. 6. Use relative paths — all file references relative to repo root. 7. Mark customization points — clearly indicate which values, files, or patterns the reader is expected to change vs. which are framework/infrastructure. 8. Version-pin everything — mention specific versions of dependencies that are known to work.
💡 Usage Examples
With Claude or ChatGPT
[Paste the prompt above]
Now scan this repository and generate the prometheus.md:
[paste your repository URL or key files]
With an MCP-enabled agent (filesystem access)
[Paste the prompt above]
The repository is located at /path/to/repo. Start by listing the directory structure,
then read the key files, and generate prometheus.md.
With GitHub MCP
[Paste the prompt above]
The repository is at github.com/user/repo. Clone it and scan all files
to produce the prometheus.md guide.
📚 What is prometheus.md?
A prometheus.md file is a comprehensive guide that lives in your repository and helps AI agents
understand how to work with your code. It's designed to be:
Agent-friendly: Written with precise file paths, line numbers, and exact commands
Comprehensive: Everything needed to understand, use, and extend your project in one place
Maintainable: Single source of truth that evolves with your codebase
Place your prometheus.md in the root of your repository, and Prometheus will automatically index it
when you connect your GitHub repository.
Ready to get started? Sign up to create and manage your guides with Prometheus.