JeffreysPrompts.com

A curated collection of battle-tested prompts for agentic coding

---

"Where'd you get that prompt?! It's lovely!"
"Oh, this old thing? I got it at JeffreysPrompts.com."

---

Browse. Copy. Export markdown. Ship faster.

What is JeffreysPrompts.com?

JeffreysPrompts.com is a platform for discovering, copying, exporting, and operationalizing curated prompts for AI coding agents like Claude Code, Codex CLI, and Gemini CLI.

At the repo level, it's four closely related layers:

Claude Code skills are managed separately in Jeffrey's Skills.md (jsm). JFP focuses on prompts and markdown export.

The current codebase also includes roadmap, ratings, review, sharing, referral, history, status, admin, moderation, appeals, DMCA, and support-ticket surfaces around the core prompt experience.

┌─────────────────────────────────────────────────────────────────────────┐
│                                                                         │
│   YOU                    jeffreysprompts.com                   AGENT    │
│    │                            │                                  │    │
│    ├─── Browse prompts ────────►│                                  │    │
│    │                            │                                  │    │
│    ├─── Copy/export ───────────►│────── jfp export ───────────────►│    │
│    │                            │            prompt.md             │    │
│    │◄───── Agent uses prompt ───│◄─────────────────────────────────│    │
│                                                                         │
└─────────────────────────────────────────────────────────────────────────┘

The Prompts

These prompts come from Jeffrey Emanuel's "My Favorite Prompts" series on Twitter. They're battle-tested patterns refined through extensive real-world usage with AI coding agents.

The Idea Wizard

Generate 30 improvement ideas, rigorously evaluate each, distill to the very best 5

The key insight: by forcing the agent to generate many ideas and then critically evaluate them, you get much better results than asking for "5 good ideas" directly.

The README Reviser

Update documentation for recent changes, framing them as "how it always was"

Addresses documentation drift. The framing trick produces cleaner, more professional docs.

The Robot-Mode Maker

Create an agent-optimized CLI for any project

Builds what the agent would want to use, because it WILL be using it. JSON output, token efficiency, quick-start mode.

...and more. Each prompt includes when to use it, tips, and examples.

Table of Contents

• Quick Start
  • Web App
  • CLI Tool
• Features
  • Web App Features
  • CLI Features
• The jfp CLI
  • Installation
  • Commands
  • Robot Mode
• Architecture
  • TypeScript-Native Prompts
  • Monorepo Layers
  • Project Structure
  • API Surface
• Development
  • Prerequisites
  • Setup
  • Quality Gates
• The Making-Of Page
• Design Philosophy
• Contributing
• License

Quick Start

Web App

Visit jeffreysprompts.com to:

1. Browse all prompts with search and filtering
2. Click any prompt to copy it to your clipboard
3. Add prompts to your basket for bulk export
4. Download as markdown

CLI Tool

curl -fsSL "https://jeffreysprompts.com/install-cli.sh?$(date +%s)" | bash

Usage:

# List all prompts
jfp list

# Fuzzy search
jfp search "brainstorm"

# View a prompt
jfp show idea-wizard

# Copy to clipboard
jfp copy idea-wizard

# Interactive mode (fzf-style)
jfp i

Skills (JSM)

Claude Code skills are managed in Jeffrey's Skills.md (jsm). Use JFP for prompts and markdown export; use JSM for skills management.

Features

Web App Features

CLI Features

CLI: Free vs Premium (Paid)

The jfp CLI is open source and builds without any Pro account. Pro is only required to authenticate and access private/premium cloud data.

Premium features require an active Pro subscription ($10/month) at: https://pro.jeffreysprompts.com/pricing

If you are not subscribed, Pro commands still exist but will return a clear upgrade hint. The CLI remains fully usable for the public registry without login.

Free (no subscription required):

• jfp list, jfp search, jfp show, jfp copy, jfp export, jfp render, jfp suggest
• jfp bundles, jfp bundle, jfp categories, jfp tags, jfp random
• jfp impact, jfp graph export, jfp serve, jfp config
• jfp status, jfp refresh, jfp open, jfp doctor, jfp about, jfp completion, jfp update-cli
• jfp i (interactive browser)

Premium (paid Pro required):

• jfp login, jfp logout, jfp whoami
• jfp save, jfp notes, jfp collections, jfp sync
• jfp recommend
• jfp packs
• jfp cost
• jfp tags suggest, jfp dedupe scan
• jfp list --mine, jfp search --mine, jfp list --saved, jfp search --saved, jfp search --all

Build (free):

• bun install
• bun run jfp.ts list
• bun build --compile ./jfp.ts --outfile jfp

The jfp CLI

CLI Installation

Quick install:

curl -fsSL "https://jeffreysprompts.com/install-cli.sh?$(date +%s)" | bash

Manual install:

Download for your platform from GitHub Releases:

Linux x64:

curl -L https://github.com/Dicklesworthstone/jeffreysprompts.com/releases/latest/download/jfp-linux-x64 -o ~/.local/bin/jfp
chmod +x ~/.local/bin/jfp

macOS ARM:

curl -L https://github.com/Dicklesworthstone/jeffreysprompts.com/releases/latest/download/jfp-darwin-arm64 -o ~/.local/bin/jfp
chmod +x ~/.local/bin/jfp

CLI Commands

jfp                           # Quick-start help
jfp help                      # Full documentation
jfp list                      # List all prompts
jfp list --category ideation  # Filter by category
jfp list --tag ultrathink     # Filter by tag
jfp list --json               # JSON output for agents

jfp search <query>            # Fuzzy search
jfp search "robot" --json     # JSON output

jfp show <id>                 # Show full prompt
jfp show idea-wizard --json   # JSON output
jfp show idea-wizard --raw    # Just the prompt text

jfp random                    # Pick a random prompt
jfp random --category testing # Filtered random prompt

jfp copy <id>                 # Copy to clipboard
jfp export <id>               # Export as markdown
jfp export <id> --format md   # Export as markdown (explicit)
jfp render <id>               # Fill prompt variables/context
jfp suggest "<task>"          # Task-based recommendations

jfp status                    # Registry cache status
jfp refresh                   # Refresh cached registry
jfp config list               # Show CLI configuration
jfp config get registry.url   # Read a config key

jfp recommend                 # Personalized recommendations (Pro)
jfp recommend idea-wizard     # Related prompts (Pro)
jfp cost idea-wizard          # Estimate tokens + cost (Pro)
jfp cost --alerts             # Show recent budget alerts (Pro)

jfp login                     # Authenticate with JeffreysPrompts Pro
jfp logout                    # Clear local credentials
jfp whoami                    # Show current account
jfp save <id>                 # Save a prompt to your library (Pro)
jfp notes <id> --add "..."    # Add personal notes (Pro)
jfp sync                      # Sync premium library locally (Pro)
jfp collections               # Manage collections (Pro)

jfp impact <id>               # Show downstream dependencies
jfp graph export              # Export prompt dependency graph

jfp packs --installed         # List installed premium packs (Pro)
jfp packs install <id>        # Install/subscribe to premium packs (Pro)
jfp packs update <id>         # Update a premium pack (Pro)
jfp packs changelog <id>      # View a pack changelog (Pro)

jfp i                         # Interactive browser (fzf-style)
jfp serve                     # Start MCP server on stdio
jfp serve --config            # Print Claude Desktop MCP config snippet

jfp categories                # List categories
jfp tags                      # List tags with counts
jfp tags suggest <id>         # Suggest tags/categories (Pro)
jfp dedupe scan               # Scan for duplicate prompts (Pro)

jfp completion --shell zsh    # Generate shell completion script

jfp --version
jfp --help

Robot Mode

The CLI is designed agent-first. When an AI coding agent uses jfp:

TTY Detection:

# Human in terminal — gets pretty output
jfp search "brainstorm"

# Agent piping output — automatically gets JSON
results=$(jfp search "brainstorm")
echo $results | jq '.results[0].id'

Structured Errors:

{
  "error": "prompt_not_found",
  "message": "No prompt with id 'foo-bar'",
  "suggestions": ["foo-baz", "idea-wizard"],
  "exitCode": 1
}

Exit Codes:

Quick Start Output (~100 tokens):

jfp — Jeffrey's Prompts CLI

QUICK START:
  jfp list                    List all prompts
  jfp search "idea"           Fuzzy search
  jfp show idea-wizard        View full prompt
  jfp export idea-wizard      Export as markdown

ADD --json TO ANY COMMAND FOR MACHINE-READABLE OUTPUT

EXPLORE:
  jfp i                       Interactive browser (fzf-style)

MORE: jfp help | Docs: jeffreysprompts.com

Architecture

TypeScript-Native Prompts

Prompts are defined as TypeScript objects, not markdown files:

// packages/core/src/prompts/registry.ts

export const prompts: Prompt[] = [
  {
    id: "idea-wizard",
    title: "The Idea Wizard",
    description: "Generate 30 ideas, rigorously evaluate, distill to best 5",
    category: "ideation",
    tags: ["brainstorming", "improvement", "ultrathink"],
    author: "Jeffrey Emanuel",
    twitter: "@doodlestein",
    version: "1.0.0",
    featured: true,
    difficulty: "intermediate",
    estimatedTokens: 500,
    created: "2025-01-09",
    content: `Come up with your very best ideas for improving this project...`,
    whenToUse: [
      "When starting a new feature or project",
      "When reviewing a codebase for improvements",
    ],
    tips: [
      "Run this at the start of a session for fresh perspective",
      "Combine with ultrathink for deeper analysis",
    ],
  },
];

Why TypeScript-native?

Monorepo Layers

The repo is organized around a shared data/model layer with multiple consumers:

The web app prebuild step validates prompt data and generates registry.json / registry.manifest.json from packages/core, keeping the published registry aligned with the typed source data.

Project Structure

jeffreysprompts.com/
├── README.md
├── AGENTS.md                      # Rules for AI agents
├── PLAN_TO_MAKE_...md             # Detailed implementation blueprint
├── Cargo.toml                     # Rust workspace for jfp port
├── package.json
├── bun.lock
│
├── jfp.ts                         # CLI entrypoint
├── jfp.test.ts                    # CLI tests
├── crates/
│   └── jfp/                       # Rust CLI port
├── packages/
│   ├── core/                      # Shared prompt data, search, export, templates
│   └── cli/                       # TypeScript CLI commands and helpers
│
├── .claude/
│   └── skills/                    # Internal authoring helpers (not shipped)
│       ├── prompt-formatter/
│       │   └── SKILL.md
│       └── skill-maker/
│           └── SKILL.md
├── apps/
│   └── web/                       # Next.js 16 App
│       ├── src/
│       │   ├── app/               # App Router pages
│       │   ├── components/        # React components
│       │   ├── hooks/             # Client hooks for filters, ratings, history, etc.
│       │   └── lib/               # Web-only stores and service modules
│       └── package.json
│
└── scripts/
    ├── build-cli.sh               # Build jfp binaries
    ├── build-data.ts              # Generate registry JSON assets
    ├── validate-prompts.ts        # Validate prompts/bundles/workflows
    └── extract-transcript.ts      # Extract Claude Code session

API Surface

The Next.js app exposes more than just prompt listing endpoints. Current route groups include:

• /api/prompts, /api/export, /api/featured, /api/skills/[id]
• /api/ratings, /api/reviews, /api/share, /api/history
• /api/roadmap, /api/referral/*, /api/status/*, /api/health/*
• /api/admin/*, /api/appeals, /api/dmca, /api/support/tickets

Those APIs back public browsing features plus internal/admin workflows that live in the same monorepo.

Development

Prerequisites

• Bun 1.x — JavaScript runtime and package manager
• Node.js 20+ — For Next.js compatibility
• Git — Version control

Setup

# Clone the repository
git clone https://github.com/Dicklesworthstone/jeffreysprompts.com.git
cd jeffreysprompts.com

# Install dependencies
bun install

# Start development server
cd apps/web
bun run dev

# Run CLI in development
bun run jfp.ts list

# Build CLI binary
bun build --compile ./jfp.ts --outfile jfp

Quality Gates

# Web app
cd apps/web
bun run test          # Unit tests (vitest + happy-dom)
bun run build         # Production build
bun run lint          # ESLint
bun run lint:all      # ESLint + Oxlint

# CLI
bun run test:cli

Important: Always use bun run test, never bun test. The latter bypasses vitest's DOM environment configuration.

The Making-Of Page

Visit jeffreysprompts.com/how_it_was_made for a unique meta-feature:

The complete, unedited Claude Code session transcript that designed, planned, and implemented this entire site — in a single day.

This page includes:

• Session Statistics — Duration, messages, tool calls, files edited, lines written
• Visual Timeline — Every message with expandable tool calls
• Syntax Highlighting — Code shown with proper highlighting
• Annotations — Commentary on key decisions and patterns
• Extended Thinking — Claude's reasoning process exposed

Why this exists:

1. Transparency — AI-assisted development shouldn't be a black box
2. Education — Learn prompt engineering from hundreds of real examples
3. Meta-demonstration — The prompts on this site helped build this site

Design Philosophy

1. Agent-First, Human-Compatible

Every feature considers the AI agent user first:

• --json flag on every CLI command
• Quick-start mode is token-dense (~100 tokens)
• Predictable response shapes
• Meaningful exit codes

2. Copy Is King

The primary action is copying prompts:

• Copy button prominent on every card
• One-click with visual feedback
• Bulk export for collections

3. Prompts First, Skills Elsewhere

JFP is for prompt discovery, copying, and markdown export:

• Clean, shareable markdown export
• Clear boundary: prompts in JFP, skills in JSM
• Agent-friendly CLI output

4. TypeScript-Native Data

No markdown files, no parsing:

• Type safety catches errors at compile time
• IDE autocomplete for all fields
• Single source of truth

5. Mobile Excellence

Following brennerbot.org patterns:

• Touch targets are 44px minimum
• Bottom sheet for mobile interactions
• No horizontal scroll
• iOS Safari visual viewport fixes

Contributing

This project welcomes contributions, but please note:

The prompts themselves are curated by Jeffrey Emanuel. If you have a prompt suggestion, open an issue to discuss it first.

Bug fixes, performance improvements, and documentation enhancements are always welcome via pull request.

Before contributing:

1. Read AGENTS.md for project conventions
2. Read PLAN_TO_MAKE_JEFFREYSPROMPTS_WEBAPP_AND_CLI_TOOL.md for architecture context
3. Run quality gates before submitting

License

MIT License (with OpenAI/Anthropic Rider). See LICENSE for details.