← Pro Prompt Pack

Getting started

Pro Prompt Pack

You can read this before you buy. After checkout, unzip lilaforge-pro-prompt-pack. There is no installer and no server. The first session is one prompt with its variables filled in.

A free selection of the prompts is at prompts.lilaforge.com. The zip is the full set of files.

What you need installed

• A chat or coding agent you already use. Each prompt in the pack has a model note saying what kind of model it suits.
• Claude Code or Cursor, only if you want the slash commands. The files are in commands/claude-code/ and commands/cursor/.
• No Node install, no API key, and no account on this site to use the files.

The first 15 minutes

1. Read the short README

   Unzip and open README.md. prompts/ has one markdown file per category (vibe-coding, local-llm, marketing, agents, debugging, writing, code-review, research, gaming) plus prompts/all-prompts.json with the same prompts.

2. Fill one prompt and paste it

   Open prompts/debugging.md, or any other category file. Replace every {{variable}} with your details, then paste the prompt into a new chat. The README says to fill them all in before you run it.

3. Run one workflow in order

   Open workflows/feature-build.md. Run the steps in order in your chat or agent. That workflow says to keep the approval pauses. The other workflows are bug-hunt, code-quality, launch-week, pre-launch, local-rag, agent-setup, research-report, content-engine and game-content. The .json file next to each markdown file is the same steps as data. Nothing in the pack runs that JSON for you.

4. Optional: install the slash commands

   Claude Code: copy the markdown files from commands/claude-code/ into .claude/commands/ in a repo, or into your user commands directory for every project. Cursor: copy commands/cursor/ into .cursor/commands/ and run them from chat with /. The file name is the command: plan.md is /plan. The twelve are /plan, /review, /rootcause, /tests, /security, /commit, /prdesc, /docs, /explain, /cleanup, /shipcheck and /agentsmd.

5. On a small local model, start with the short prompts

   Use prompts/local-llm.md. The README says those prompts are short and strict because small 1B-9B models follow simple, explicit formats, and it says to use temperature 0–0.3 for extraction and classification.

Common problems

The model asks what a brace word means, or it invents a value

A {{variable}} is still in the prompt. The README says leaving one in usually makes the model ask or guess. Replace every one, including the ones inside a workflow step.

A slash command does not appear

The markdown is still inside the zip. Claude Code reads .claude/commands/. Cursor reads .cursor/commands/ and you run them from chat with /. Copy the files from commands/claude-code/ or commands/cursor/, not the parent folder alone. The command name is the file name without .md.

A small local model ignores a long prompt

Use the prompts in prompts/local-llm.md. The other categories are written for models that can hold a longer instruction. The model note at the top of each prompt says what it suits.

I edited the markdown and the JSON still has the old text

prompts/all-prompts.json is a second copy of the same prompts, for a script or a notes tool. The workflow JSON files are the same kind of copy. Editing the markdown does not rewrite the JSON.

See it running live: https://prompts.lilaforge.com

Back to Pro Prompt Pack for what's included and checkout.