AllanZyne

Pi Agent Views — Design skill for Claude Code

Design community

Claude Code-style agent views for pi — manage multiple agents per session.

How to install Pi Agent Views

This entry records only its repository, not the path inside it, so there is no exact command to give. Open AllanZyne/pi-agent-views and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Pi Agent Views does

Claude Code-style agent views for pi — manage multiple agents per session.

Alternatives in Design

Brand Guidelines — Apply Anthropic's official brand colors and typography to artifacts 94.1k ★
Planning With Files — Claude Code skill implementing Manus-style persistent markdown planning — the workflow pattern behind the $2B 17.4k ★
Explanatory Output Style — Educational output mode with insights about implementation choices 14k ★

README

pi-agent-views

Claude Code-style **Agent Views** for [pi](https://pi.dev) — manage multiple agents per session.

Features

Per-session agent management: Each pi session has its own agent list. The session itself is the "Main" agent; additional agents are sub-sessions stored separately from normal session management (/resume won't see them).
Background execution: Agents created via /agent run in background pi subprocesses. Switch between agents freely — running agents automatically continue in the background.
Grouped by state: Agents are displayed grouped as Working → Failed → Idle → Completed with colored icons.
Context-aware dispatch: /agent serializes the current conversation and starts a background agent that extracts relevant context and works on the task — all without blocking.
Real-time refresh: Agent Views auto-refreshes every 3 seconds while open so you can watch Working → Completed transitions.

Install

pi install git:github.com/AllanZyne/pi-agent-views

Or for project-local:

pi install -l git:github.com/AllanZyne/pi-agent-views

Usage

Open Agent Views

Inside Agent Views

Key	Action
`↑/↓`	Navigate agents (when editor is empty)
`Enter` (empty)	Attach to selected agent
`Enter` (with text)	Dispatch a new agent with that text
`→`	Attach to selected agent
`←` / `Esc`	Close Agent Views
`Space`	Toggle peek panel
`?`	Show keyboard help

Create agents

**From Agent Views** — type a prompt in the editor and press `Enter`:

fix the flaky test in auth_test.go

A new agent session is created and you switch into it.

**From any agent** — use `/agent` to create a background agent with LLM-curated context:

/agent based on our analysis, write a fix PR

The current conversation is serialized and sent to a background pi subprocess. The LLM extracts relevant context and works on the task — without block