pi-skill-loader

Codex-style skill loading for pi - type # to load skills into context without consuming an LLM turn

Packages

Package details

extension

Install pi-skill-loader from npm and Pi will load the resources declared by the package manifest.

npm repo home report

$ pi install npm:pi-skill-loader

Package: pi-skill-loader
Version: 0.1.2
Published: Oct 22, 2026
Downloads: not available
Author: wildermoth
License: MIT
Types: extension
Size: 21.4 KB
Dependencies: 0 dependencies · 1 peer

Pi manifest JSON

{
  "extensions": [
    "./index.ts"
  ]
}

Security note

Pi packages can execute code and influence agent behavior. Review the source before installing third-party packages.

README

skill-loader

Codex-style skill loading for pi: type #skillname to load a skill into context without consuming an LLM turn.

Install

pi install npm:pi-skill-loader

Or try it for one session without installing:

pi -e npm:pi-skill-loader

Usage

Type # anywhere in the editor and the skill box opens automatically, no Tab needed. Pick a skill with Tab/Enter and it is rewritten to a {$skill} marker, tinted with your theme's accent color so pending loads are visible. Esc dismisses the box and leaves plain text that will not load. You can also type the {$skill} marker by hand.

Input	Behavior
`{$skill}` (selected or typed)	Load a skill, no LLM turn consumed
`review this {$style}`	Load the skill and send the message as written
`#`	Open the skill picker menu
`/skills`	List currently loaded skills
`/skills-clear`	Reset loaded skill tracking

Only {$...} markers load. Plain #words, hashtags, and pasted markdown pass through to the agent untouched and never load. Mentions stay in your message so the reference remains visible. A message holding only skill markers loads them without consuming a turn. Each skill is submitted once as a context message when first loaded, without starting a model turn. Repeated mentions do not resubmit it. Loaded-skill tracking persists across /reload and follows the current session branch. Skill content remains in conversation history, subject to normal compaction. /skills-clear resets tracking so skills can be loaded again; it does not erase previously submitted content. Unreadable or empty skills report an error and are not marked loaded.

Highlighting replaces the editor component while installed; if another extension already owns it, highlighting is skipped with a warning and everything else keeps working.