How does the skill know when a task is successfully finished?

The skill only signals completion once all acceptance criteria defined in the initial PRD phase are met and verified during the Review step.

Can I use Autopilot specifically for bug fixes?

Absolutely. Using the --bug flag tells the autopilot to focus on reproducing the issue, applying a fix, and ensuring no regressions are introduced.

Does it work with existing project files?

Yes, it specifically leverages project files like MAP.md and TODO.md to understand your codebase structure and manage task backlogs.

How many times will the Autopilot skill attempt a task?

By default, the skill runs for a maximum of 5 iterations to allow for self-correction while preventing infinite loops.

Autopilot for Claude Code

Name: Autopilot for Claude Code
Author: atrislabs

byatrislabs

•

Productivity & Workflow

Automates the entire development lifecycle from PRD generation to final verification using a self-correcting autonomous loop.

Autopilot is a high-level orchestration skill for Claude Code that enables hands-off execution of features and bug fixes. By following a structured Plan-Do-Review loop, it automatically generates a Product Requirements Document (PRD), maps out necessary code changes, implements logic, and validates results against strict acceptance criteria. This skill is ideal for developers looking to delegate multi-step tasks to an agent that self-corrects and reports progress through a persistent journal and task tracking system, ensuring that 'done' actually means the code is tested and verified.

Key Features

01Self-Correcting Logic: Detects failures during the review phase and loops back to the planning stage for fixes.

0254 GitHub stars

03Project Context Integration: Seamlessly utilizes MAP.md and TODO.md for project-aware navigation and tracking.

04PRD-Driven Execution: Automatically creates and follows testable acceptance criteria for every task.

05Autonomous Plan-Do-Review Loop: Iteratively builds and validates code until all requirements are met.

06Structured Progress Logging: Maintains a detailed execution log and journal for full transparency of autonomous actions.

Use Cases

01Hands-off feature implementation by providing high-level prompts like 'Add dark mode toggle'.

02Clearing a backlog of tasks directly from a TODO.md file with minimal human intervention.

03Autonomous bug hunting and fixing with integrated regression testing and validation.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add atrislabs/atris autopilot

For use in Claude.ai and ChatGPT

Download Skill