todo-task

Create task plans, triage them into executable specs, and run headless agents in git worktrees.

1|Updated Jan 19, 2026
One-click install
npx skills add https://github.com/saxonthune/carta --skill todo-task
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: todo-task
Source: https://github.com/saxonthune/carta/tree/main/.claude/skills/todo-task
Command: npx skills add https://github.com/saxonthune/carta --skill todo-task

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires bash, jq, git, claude.

What problem does it solve?

This Skill removes the heavy lifting of managing an AI-assisted task lifecycle—turning rough ideas into executable implementation plans, running headless agents to make code changes, and producing a clear status/result trail.

Core Features & Use Cases

  • Unified lifecycle control: create tasks, triage them into an executable spec, execute via headless agents, and report status.
  • Deterministic execution orchestration: launches plans in isolated git worktrees, verifies changes using a fenced verification block, and merges (or leaves for manual review).
  • Robust state tracking & recovery: uses a directory-as-state-machine (.todo-tasks/ → .running/ → .done//.archived/) plus an emergency finalizer to avoid stuck runs.
  • Chain execution: run multiple triaged plans in sequence using a chain worktree, then merge the chain back into the trunk.
  • Operational dashboards: live monitoring and comprehensive status reporting with clear buckets (success, ready, questionable, attention).

Quick Start

Tell your AI to execute the next triaged task by running: "/todo-task execute add-user-search" and then watch progress via the status output.

Frequently Asked Questions about todo-task

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automate task management for codebase changes using headless agents?

Automated task management creates triaged executable specifications from rough plans and runs headless agents to implement codebase changes. It uses git worktrees for isolation and fenced bash verification to ensure deterministic, repeatable edits before merging results.

Can I execute multiple plans in sequence with agent orchestration?

Yes, agent orchestration supports chain execution to run multiple triaged plans in sequence. It uses a dedicated chain worktree to track progress and merges the completed sequence back into the trunk branch safely.

Do I need git worktrees to run isolated agent code edits?

Yes, git worktrees are required to isolate agent code edits. Worktree isolation ensures deterministic execution by keeping ongoing changes separate, preventing conflicts in the main repository until verification passes and a safe merge is initiated.

What is the best way to track task lifecycle status and recover from stuck runs?

Task lifecycle status is tracked using a directory-as-state-machine moving tasks from .todo-tasks to .running and .done. An emergency finalizer automatically recovers stuck runs, while operational dashboards report progress across success, ready, questionable, and attention buckets.

How does merge handling work for automated git worktree tasks?

Merge handling for automated git worktree tasks safely integrates verified changes back into the trunk. It includes explicit conflict handling and rollback behaviors to prevent broken states if an agent's code edits fail verification or produce invalid results.