Prompt Templates

Add model, skill, and thinking frontmatter to Pi prompt templates so slash commands switch modes automatically.

What it does

Pi-prompt-template-model extends Pi's built-in prompt templates with frontmatter like model, skill, and thinking. That lets a template become a reusable agent mode: run a single slash command and Pi can switch to the right model, preload a skill, adjust thinking depth, and restore your previous session settings when the task finishes.

The extension also supports more advanced template workflows like fallback model lists, chain templates, looping, and optional delegated execution through sub-agents. You still author normal prompt templates — this package just makes them much more powerful.

Why it's included

One of the easiest ways to make Pi feel personal is to create task-specific slash commands. Prompt Templates turns those commands into reliable workflows instead of fragile text snippets. You can make a /deep-debug template that always uses a stronger model, a /quick-fix template that stays cheap and fast, or a review template that injects exactly the right skill every time.

It fits LazyPi's philosophy well: remove repetitive setup, make good workflows easy, and let users keep the speed of Pi's minimal interface while gaining much richer behavior underneath.

Usage notes

After installation, restart Pi so the extension loads. Then add frontmatter to one of your prompt templates:

---
description: Debug Python in tmux REPL
model: claude-sonnet-4-20250514
skill: tmux
thinking: medium
---
Start a Python REPL session and help me debug: $@

When you run that template's slash command, Pi switches to the requested model, loads the skill context, and applies the chosen thinking level for the duration of the task. By default it restores your previous session settings when the command completes.

If you use delegated execution features like subagent or inheritContext, LazyPi already installs Sub-agents for you — so those workflows are available out of the box.