Scriptify

Crystallize and reuse agent actions by preserving agent produced disposable scripts.

Why?

Agents constantly write throwaway scripts to answer our requests. Ask one "which pods restarted the most in the last 24h?" and it will likely write a small program, run it, show you the result, and throw the program away.

Across day-to-day engineering work, that adds up to a large body of scripts that are lost when they could be:

• Reused outside the agent: dropped into cron jobs, CI, or other deterministic automation.
• Trusted: a saved script has been reviewed and behaves the same way every time, unlike a fresh regeneration that may differ in small ways.
• Recalled instead of regenerated: writing the same script again costs tokens, time, energy, and money.

Scriptify fixes this in two ways. It saves useful scripts as parameterized, reusable tools that live outside the session. And it makes the agent recall them from that library before writing new code, so it stops reinventing the wheel.

Installation

As a Claude Code plugin

/plugin marketplace add pfcoperez/scriptify
/plugin install scriptify@scriptify

The plugin bundles the skill and a Stop hook that makes it trigger reliably (see below). The hook requires python3.

For other agents

npx skills add https://github.com/pfcoperez/scriptify

Or just ask your agent to install it.

Prerequisites

• An AI coding agent with npx skills support
• Python3, if installed as Claude Code plugin.

Usage

The skill is designed to alter agent behavior without active user request but it can be invoked manually with:

/scriptify

📝If not used as a Claude Code plugin, the invocation of the skill is not deterministic. In these cases, AGENTS.md can be used to increase the chances of it being invoked by adding instructions such as:

Invoke /scriptify after you have finished working on a prompt