Agent skill

Pi Subagent

by espennilsen in espennilsen/pi

Delegate tasks to specialized subagents for parallel or sequential execution.

MITAuto-check passedAgent Workflows

Install Pi Subagent

skills CLI
$ npx skills add espennilsen/pi --skill pi-subagent -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install espennilsen/pi pi-subagent --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/espennilsen/pi.git skills-src && mkdir -p .claude/skills && cp -r skills-src/extensions/pi-subagent/skills/pi-subagent .claude/skills/pi-subagent && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
pi-subagent
GitHub stars
122
Token cost
~1.2k tokens
SKILL.md length
364 words
Files
1
Skills in repo
36
Repo updated
First seen
Licence
MIT

At a glance

Delegate tasks to specialized subagents for parallel or sequential execution.

  • Tasks that involve Subagents
  • SKILL.md covers When to use subagents, Modes, Per-task options and Extension isolation, plus 3 more sections
  • Instructions only: no scripts, shell commands, URLs or credentials in SKILL.md

What it does

Pi Subagent is an agent skill from espennilsen/pi. Delegate tasks to specialized subagents for parallel or sequential execution.

Its SKILL.md is about 1.2k tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.

It sits in Agent Workflows, covering Subagents. The licence is MIT.

When your agent uses it

  • Tasks that involve Subagents

Example prompts

  • “/pi-subagent”

What it can do on your machine

Read from SKILL.md and the folder at commit 79d019b. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    No scripts in the folder and no shell commands in SKILL.md (its code samples are json and yaml).

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Pi Subagent loads about 1.2k tokens when it runs. Until then it costs about 22 tokens; SKILL.md has 364 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~22
When it runs · the whole SKILL.md, loaded when a task matches
~1.2k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from espennilsen/pi at commit 79d019b, republished under its MIT licence (© espennilsen). 364 words, ~1,177 tokens.

Download SKILL.mdSave it as .claude/skills/pi-subagent/SKILL.md (or your agent's skills folder).
name
pi-subagent
description
Delegate tasks to specialized subagents for parallel or sequential execution.

Subagent Orchestration

Use the subagent tool to delegate tasks to isolated pi subprocesses. Each subagent gets a fresh context window — no shared state with the parent session.

When to use subagents

  • A task can be split into independent pieces that run in parallel
  • You need a chain of specialists (scout → planner → worker)
  • A subtask benefits from a clean context window (no accumulated noise)
  • You want crash isolation (failed subtask doesn't corrupt main session)

Modes

Single — one agent, one task
json
{ "agent": "scout", "task": "Find all REST endpoints in " }
Parallel — multiple agents concurrently (streams progress as each completes)
json
{ "tasks": [
    { "agent": "scout", "task": "Audit auth module" },
    { "agent": "scout", "task": "Audit API routes" },
    { "agent": "scout", "task": "Audit database queries" }
  ]
}
Chain — sequential pipeline, use {previous} for prior step's output
json
{ "chain": [
    { "agent": "scout", "task": "Find all TODO comments in the codebase" },
    { "agent": "planner", "task": "Categorize and prioritize these TODOs:\n{previous}" },
    { "agent": "worker", "task": "Fix the top 3 critical TODOs:\n{previous}" }
  ]
}

Per-task options

Every task (single, parallel item, chain step) supports these overrides:

OptionTypeDescription
modelstringModel override (e.g. "claude-haiku-4-5" for fast, "claude-sonnet-4-5" for complex)
thinkingstringThinking level: off, minimal, low, medium, high, xhigh
extensionsstring[]Extension paths to load (subagents run with -ne — only whitelisted)
skillsstring[]Skill files/dirs to load via --skill
noToolsbooleanDisable all built-in tools (--no-tools) — for analysis-only agents
noSkillsbooleanDisable skill discovery (-ns)
cwdstringWorking directory override

Priority: per-task item > top-level params > agent .md frontmatter > global settings

Examples

Fast recon with cheap model, no thinking overhead:

json
{ "agent": "scout", "task": "Map the auth module", "model": "claude-haiku-4-5", "thinking": "off" }

Research task needing web search:

json
{ "agent": "worker", "task": "Find pricing for Vercel Pro", "extensions": ["extensions/pi-brave-search"] }

Parallel with different models per task:

json
{ "tasks": [
    { "agent": "scout", "task": "Map the codebase", "model": "claude-haiku-4-5" },
    { "agent": "reviewer", "task": "Review the PR diff", "model": "claude-sonnet-4-5", "thinking": "high" }
  ]
}

Chain with escalating capability:

json
{ "chain": [
    { "agent": "scout", "task": "Map the auth module", "model": "claude-haiku-4-5", "thinking": "off" },
    { "agent": "planner", "task": "Plan a refactor based on: {previous}", "thinking": "high" },
    { "agent": "worker", "task": "Implement the plan: {previous}", "thinking": "medium" }
  ]
}
Show full SKILL.md (156 more words)Show less

Extension isolation

Subagents always run with --no-extensions (-ne). They cannot:

  • Spawn further subagents (pi-subagent is always blocked)
  • Send messages via channels, cron, or heartbeat
  • Access the web dashboard or webserver

Whitelist only what a specific task needs via the extensions parameter.

Agent scope

  • "user" (default) — loads from ~/.pi/agent/agents/*.md
  • "both" — also includes .pi/agents/*.md (prompts for confirmation)
  • "project" — only project-local agents

Creating agents

Place .md files in ~/.pi/agent/agents/ with YAML frontmatter:

yaml
---
name: scout
description: Fast codebase reconnaissance
tools: read, grep, find, ls, bash
extensions: extensions/pi-dotenv
model: claude-haiku-4-5
---
System prompt for the agent goes here...
FrontmatterDescription
nameAgent name (required)
descriptionWhat the agent does (required)
toolsComma-separated tool whitelist
modelDefault model
extensionsComma-separated extension paths to load

Tips

  • Scout first: use scout (haiku, fast) for recon, then hand results to planner/worker
  • Parallel for independence: only parallelize truly independent tasks
  • Chain for pipelines: scout → planner → worker is the classic pattern
  • Model matching: use haiku for simple recon, sonnet for analysis/implementation
  • Thinking levels: off for fast tasks, high/xhigh for complex reasoning
  • Keep extensions minimal: only whitelist what's actually needed per task

© espennilsen, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in extensions/pi-subagent/skills/pi-subagent of espennilsen/pi.

Open the folder on GitHubat commit 79d019b

Compare with similar skills

Pi Subagent next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Pi Subagent compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Pi Subagent this skillespennilsen/pi122—~1.2kAutomated safety check: PassMIT
Claude Code Agent Developmentanthropics/claude-plugins-official38k7 repos~2.8kAutomated safety check: PassApache-2.0
Subagent Driven DevelopmentAsvarox/allkaraoke26137 repos~1.2kAutomated safety check: PassNone
Dispatching Parallel Agentsultralisp/ultralisp25840 repos~1.5kAutomated safety check: PassNone
Paseo Advisor Second Opiniongetpaseo/paseo20k1 repos~756Automated safety check: PassCustom licence
Task Observerrebelytics/one-skill-to-rule-them-all3.2k1 repos~12kAutomated safety check: PassCC-BY-4.0

Similar skills

  • Claude Code Agent Development

    anthropics/claude-plugins-official

    Official

    Explains how to write agents for Claude Code plugins: the markdown file with YAML frontmatter, trigger descriptions, model and color settings, and system prompt design.

    38k GitHub starsUsed in 7 repos~2.8k tokens
    Agent WorkflowsAuto-check passed
  • Subagent Driven Development

    Asvarox/allkaraoke

    A skill your agent uses when executing implementation plans with independent tasks in the current session

    261 GitHub starsUsed in 37 repos~1.2k tokens
    Agent WorkflowsAuto-check passed
  • Dispatching Parallel Agents

    ultralisp/ultralisp

    A skill your agent uses when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies

    258 GitHub starsUsed in 40 repos~1.5k tokens
    Agent WorkflowsAuto-check passed
  • Launches one separate agent through Paseo to give a second opinion on the current task, with a self-contained briefing and no permission to edit files.

    20k GitHub starsUsed in 1 repo~756 tokens
    Agent WorkflowsAuto-check passed
  • Task Observer

    rebelytics/one-skill-to-rule-them-all

    Monitors task execution for skill improvement opportunities.

    3.2k GitHub starsUsed in 1 repo~12k tokens
    Agent WorkflowsAuto-check passed
  • O2 Review Loop

    openobserve/openobserve

    Splits a change into planner, coder and independent reviewer roles: you confirm a spec, a subagent implements it, and a separate reviewer checks each round's local WIP commit.

    22k GitHub stars~3.7k tokensUpdated today
    Agent WorkflowsAuto-check passed

More from espennilsen/pi

All 36 skills in this repo
  • GitHub

    espennilsen/pi

    Interact with GitHub repos, PRs, issues, CI, and notifications via the pi-github extension commands and gh CLI.

    122 GitHub stars~1k tokensUpdated 17 days ago
    Auto-check passed
  • Skill Creator

    espennilsen/pi

    Create, review, and improve skills for Pi agents. An agent skill from espennilsen/pi.

    122 GitHub stars~2.1k tokensUpdated 17 days ago
    Auto-check passed
  • Dry Code Review

    espennilsen/pi

    Perform a comprehensive DRY (Don't Repeat Yourself) code review on a codebase.

    122 GitHub stars~1.7k tokensUpdated 17 days ago
    Auto-check passed
  • Extract Design System

    espennilsen/pi

    Reverse-engineer a design system from a live website (public URL or localhost).

    122 GitHub stars~2k tokensUpdated 17 days ago
    Auto-check passed
  • PDF Reader

    espennilsen/pi

    Read and extract content from PDF files — text, tables, metadata, and images.

    122 GitHub stars~1.6k tokensUpdated 17 days ago
    Auto-check passed
  • Google Workspace

    espennilsen/pi

    Manage Google Workspace via the gws CLI — Drive, Gmail, Sheets, Docs, Slides, People, Chat, Meet, Forms, and cross-service workflows.

    122 GitHub stars~3.4k tokensUpdated 17 days ago
    Auto-check passed

Categories

Questions about Pi Subagent

What does Pi Subagent do?

Delegate tasks to specialized subagents for parallel or sequential execution. Pi Subagent is an agent skill from espennilsen/pi. Delegate tasks to specialized subagents for parallel or sequential execution.

When should I use Pi Subagent?

Pi Subagent fits situations like: tasks that involve Subagents.

How do I install Pi Subagent in Claude Code?

Run `npx skills add espennilsen/pi --skill pi-subagent -a claude-code`. Or copy the skill folder (extensions/pi-subagent/skills/pi-subagent in espennilsen/pi) into .claude/skills/pi-subagent in your project. Claude Code loads it when a task matches its description.

How do I install Pi Subagent in Codex?

Run `npx skills add espennilsen/pi --skill pi-subagent -a codex`. Or copy the skill folder (extensions/pi-subagent/skills/pi-subagent in espennilsen/pi) into .agents/skills/pi-subagent in your project. Codex loads it when a task matches its description.

Can I use Pi Subagent in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add espennilsen/pi --skill pi-subagent -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/pi-subagent, .gemini/skills/pi-subagent, .github/skills/pi-subagent and .opencode/skills/pi-subagent in your project.

What does Pi Subagent need to run?

SKILL.md names no scripts, command-line tools or credentials: Pi Subagent is instructions for the agent only.

Does Pi Subagent access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Pi Subagent safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Pi Subagent use?

Pi Subagent is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Pi Subagent use?

About 1.2k tokens (SKILL.md is roughly 4.7k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Pi Subagent?

Skills that share tags, products or a category with Pi Subagent: Claude Code Agent Development (anthropics/claude-plugins-official, 38k stars), Subagent Driven Development (Asvarox/allkaraoke, 261 stars), Dispatching Parallel Agents (ultralisp/ultralisp, 258 stars) and Paseo Advisor Second Opinion (getpaseo/paseo, 20k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Pi Subagent?

espennilsen (a GitHub user) maintains it in espennilsen/pi, which has 122 GitHub stars. The repository holds 36 skills in this directory. The repository was last updated on September 21, 2026.

Source: espennilsen/pi on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.