pi-pinata

Fast, low-footprint Pi subagents: in-process agents, live-checkout readers, reviewed builder changes, and Herdr or external viewers when you want them.

Packages

Package details

extensionskillprompt

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

npm repo home report

$ pi install npm:pi-pinata

Package: pi-pinata
Version: 1.0.0
Published: Oct 26, 2026
Downloads: not available
Author: funsaized
License: MIT
Types: extension, skill, prompt
Size: 600.2 KB
Dependencies: 0 dependencies · 3 peers

Pi manifest JSON

{
  "skills": [
    "./skills/subagents",
    "./skills/engmgmt"
  ],
  "themes": [],
  "prompts": [
    "./prompts/scout.md",
    "./prompts/research.md",
    "./prompts/planner.md",
    "./prompts/builder.md",
    "./prompts/reviewer.md",
    "./prompts/pinata-review.md",
    "./prompts/pinata-fix.md"
  ],
  "extensions": [
    "./engine/pi/extension.ts"
  ]
}

Security note

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

README

piñata

Fast, observable subagents for Pi.

Ask for a scout, a researcher, a planner, a builder, or a reviewer. Pi delegates the work, the agents run in parallel where they can, and their results come back to Pi. Every builder change gets an independent review before it reaches your checkout.

Install

In Pi's terminal (manual):

pi install npm:pi-pinata

Reload Pi. You need Pi 1.1.0+, Node 22.19+ (for the pinata command), and Git, on Linux, macOS or Windows. Research tasks also use pi-web-access. Herdr is optional: inside Herdr, agents can run in their own panes.

Try this first

Ask Pi in plain language:

Have a scout figure out how request validation works in this repo.

Send two scouts out in parallel, one on the retry logic and one on its tests.
Then have research check both reports against the official docs.

/pinata-fix Correct the seconds-to-milliseconds conversion in src/duration.mjs.

/pinata-review Focus on the error handling.

Pi picks the agents, writes their briefs, and decides who waits for whom. Start with the first-run tutorial.

Meet the team

Agent	What it does
Scout (`scout`)	Finds where the behavior lives.
Researcher (`research`)	Checks the docs and brings sources.
Planner (`planner`)	Turns findings into a workable plan.
Builder (`builder`)	Makes the change and runs the checks.
Reviewer (`reviewer`)	Checks the diff and challenges the assumptions.

What it's like to use

Fast. Agents run inside your Pi by default: one starts in a few milliseconds and uses a megabyte or two, so many can work at once.
Visible. A widget above the editor shows every agent's state, time, tokens, cost and current tool. /pinata open <task> shows an agent's whole conversation as it streams, and you can steer it from there. pinata view follows a run from another terminal.
Reviewed before it lands. Each builder works in its own git worktree; its checks run after it finishes, and an independent reviewer must approve that exact change before pinata applies it to your checkout (never staged or committed). You can roll it back.
Recoverable. Runs are logged on disk. A run started with survive: true keeps going when Pi exits; the next Pi picks it up. /pinata rerun restarts tasks lost in a crash.
Scriptable. pinata run job.json runs a job without an interactive Pi, with plain or JSONL output and exit codes for CI.

Documentation

I want to…	Start here
Run my first agents	First run
Build, review and apply a change	Build and review
Watch agents work	Watching agents
Choose lean or observe mode	Modes
Run agents as processes or Herdr panes	Backends, Herdr
Run jobs from scripts or CI	Headless runs
Give builders their dependencies	Dependencies
Recover, rerun or clean up	Recovery
Set up on Windows or macOS	Platforms
Look up a tool, command or config key	Reference
Understand how it works and what it trusts	Architecture, Trust

The documentation index lists everything, including maintainer docs.

MIT licensed.