Prompt Runner

A simple tool to run prompts against files using various LLM models.

Installation

1. Make sure you have the llm command line tool installed
2. Clone this repository
3. Make the script executable: chmod +x run-prompt

ZSH Completion

To enable zsh completion for run-prompt:

(update to where you have this installed)

# Add this to your .zshrc
PROMPT_LIBRARY_PATH="/Users/wschenk/prompt-library"
fpath=($PROMPT_LIBRARY_PATH $fpath)
export PATH="$PROMPT_LIBRARY_PATH:$PATH"
autoload -Uz compinit
compinit

Usage

CLI

run-prompt <prompt_file> <input_file>

You can optionally specify a different model using the MODEL environment variable:

MODEL=claude-3.7-sonnet run-prompt <prompt_file> <input_file>

The default model is claude-3.7-sonnet.

MCP Server

npx @modelcontextprotocol/inspector uv run run-prompt mcp

Available Prompts

Content Analysis

• summarize.md - Generate 5 different two-sentence summaries to encourage readership
• key-themes.md - Extract key themes from the input text
• linkedin.md - Format content as an engaging LinkedIn post

Code Review

• lint.md - Assess code quality and suggest improvements
• git-commit-message.md - Generate semantic commit messages from code diffs

Repository Analysis

• architecture-review.md - Review architectural patterns and decisions
• api-documentation.md - Generate API documentation
• performance-review.md - Analyze performance considerations
• security-review.md - Review security implications
• developer-guide.md - Create developer documentation

Examples

Summarize a README file:

./run-prompt content/summarize README.md

Extract key themes from a document:

./run-prompt content/key-themes document.txt

Format content for LinkedIn:

./run-prompt content/linkedin article.txt

Generate a commit message:

./run-prompt code/git-commit-message.md diff.txt

Review code quality:

./run-prompt code/lint.md source_code.py

Adding New Prompts

Add new prompt files to the appropriate directory:

• content/ - For content analysis and formatting prompts
• code/ - For code-related prompts
• code/repomix/ - For repository analysis prompts

The prompt file should contain the instructions/prompt that will be sent to the LLM along with the content of your input file.

Usage examples

ollama

cat repomix-output.txt | ollama run gemma3:12b "$(cat ~/prompts/code/repomix/developer-guide.md )"

llm install llm-ollama

llm

MODEL=${MODEL:-claude-3.7-sonnet}

cat repomix-output.txt | \
llm -m $MODEL \
	"$(cat ~/prompts/code/repomix/developer-guide.md )"

---

Prompt Viewer PWA

This repository also includes a Progressive Web App (PWA) for browsing and copying prompts on mobile devices.

Features

• 📱 Install as a mobile app
• 🔍 Browse all prompts with folder navigation
• 📋 One-click copy to clipboard
• 🌐 Works offline
• 🌓 Auto light/dark mode

Deployment

The PWA is located in the pwa/ directory. To deploy it on GitHub Pages:

1. Go to Settings → Pages in your GitHub repository
2. Select "Deploy from a branch"
3. Choose your main branch and /pwa folder as the source
4. Save the settings

After a few minutes, your PWA will be available at:

https://the-focus-ai.github.io/prompt-library/

Using the PWA

1. Visit the URL on your mobile device
2. You'll see an "Add to Home Screen" prompt (or use browser menu)
3. Once installed, it works like a native app
4. Click refresh to cache all prompts for offline use

For more details, see pwa/deployment.md.