Use shared, persistent task lists to track multi-step work across AI agents and sessions.

Install

mkdir -p .claude/skills/pi-tasks && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/11834" && unzip -o skill.zip -d .claude/skills/pi-tasks && rm skill.zip

Installs to .claude/skills/pi-tasks

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

Use when work should be tracked in persistent shared pi-tasks lists: multi-step plans, resumable work, handoffs, parallel agents, task claiming, task status updates, or when the user asks to create, list, claim, update, or close tasks. Explains when and how to use the pi-tasks extension tools.
294 chars✓ has a “when” triggerlonger than Claude Code's old 250-char listing cap (fine on current versions)
Intermediate

Key capabilities

  • Track work in persistent shared task lists
  • Create, find, get, and delete task lists
  • Add multiple tasks to a list
  • Claim the next available task for an agent
  • Refresh a task claim for long-running work
  • Update task status and add notes for working memory

How it works

This skill manages persistent shared task lists for agents, allowing for creation, claiming, updating, and auditing of tasks. It uses time-limited locks for task claims to coordinate work.

Inputs & outputs

You give it
Task list name, scope type/key, task titles, task ID, status, outcome
You get back
Task list details, claimed task ID, updated task status, audit events

When to use pi-tasks

  • Tracking multi-step plans
  • Handoffs between agents
  • Resumable work tracking
  • Parallel agent coordination

About this skill

pi-tasks

Persistent shared task lists for Pi agents and MCP clients.

When to use

  • The user asks for task lists, todo queues, claims, assignments, handoffs, or status tracking.
  • Work has multiple steps and should survive reload, compaction, or session changes.
  • Multiple agents/sessions/MCP clients may coordinate on the same queue.
  • You need to pick the next available unit of work instead of choosing manually.

When not to use

  • A one-shot answer or tiny edit needs no durable task state.
  • A normal prose plan is enough and the user did not ask for persistent tracking.
  • A private list is inaccessible, unless the user explicitly confirms a bypass.

Minimal workflow

  • Use task_help when unsure; its Pi UI output is concise and is the source of truth.
  • Use task_lists action=find|get|create|delete to discover or manage lists.
  • Use task_items action=create|add_many to populate a list.
  • Use task_claims action=claim_next as the only normal way to move work to in_progress.
  • A claim is a time-limited lock. claim_expires_at shows when it expires.
  • Use task_claims action=refresh for long-running work so another agent cannot reclaim it.
  • After expiry, the task still exists but the lock is stale; release_expired or a later claim_next clears that stale lock so the task can be claimed again.
  • Use task_items action=update with notes for task-local working memory.
  • Close tasks with task_items action=update and status done or canceled; outcome is required.
  • Pause tasks with status blocked; omit assigned_to_agent_id to keep responsibility, pass null to release it.

Tool map

  • task_lists: create, find, get, delete lists.
  • task_items: create, add_many, update, reorder, delete tasks.
  • task_claims: claim_next, refresh, release_expired claims.
  • task_audit: read visible private-list bypass audit events.
  • task_help: show workflow, schemas, and examples.

Common calls

Find lists for the current workspace:

{ "action": "find", "params": { "scope_type": "workspace", "scope_key": "/repo" } }

Create a shared list:

{ "action": "create", "params": { "name": "Release work", "scope_type": "workspace", "scope_key": "/repo" } }

Add several tasks:

{ "action": "add_many", "params": { "list_id": "release", "tasks": [{ "title": "Run tests" }] } }

Claim next task:

{ "action": "claim_next", "params": { "list_id": "release" } }

Complete a task:

{ "action": "update", "params": { "task_id": "task-id", "status": "done", "outcome": "Decision: ship. Actions: tests. Final state: green." } }

When not to use it

  • When a one-shot answer or tiny edit needs no durable task state
  • When a normal prose plan is sufficient and persistent tracking is not requested
  • When a private list is inaccessible and the user does not explicitly confirm a bypass

Limitations

  • A claim is a time-limited lock
  • After expiry, the task still exists but the lock is stale
  • Private lists are inaccessible unless explicitly bypassed

How it compares

This skill provides a structured, persistent task tracking system with claiming and refreshing mechanisms, enabling multi-agent coordination and resumable work, unlike simple todo lists.

Compared to similar skills

pi-tasks side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
pi-tasks (this skill)03moNo flagsIntermediate
task-management03moNo flagsBeginner
yatta06moReviewIntermediate
task-execution-engine37moReviewBeginner

Try saying

Example prompts that trigger this skill in your AI assistant.

You might also like

Search skills

Search the agent skills registry