How do I start the autopilot mode?

You can start the process by typing `/task-ai:auto` or using conversational triggers like 'auto', 'run automatically', or 'start autopilot'.

What happens if the AI gets stuck in a retry loop?

The skill includes adaptive retry limits. If a plan or execution step fails to meet the quality threshold after a set number of attempts, it will stop and notify the user that manual intervention is required.

Can I interrupt the automated process once it starts?

Yes, you can stop the execution loop at any time using the `/task-ai:auto stop` command or by providing manual feedback to override the current phase.

What are the D1-D6 quality gates?

These are standardized dimensions used to evaluate the quality of deliverables against the original target and plan, ensuring the output meets specific scoring thresholds before proceeding.

Task Lifecycle Autopilot

Name: Task Lifecycle Autopilot
Author: huacheng

byhuacheng

•

生産性とワークフロー

Automates the end-to-end development lifecycle from requirement definition to final acceptance through a conversational AI workflow.

This skill provides a fully automated, hands-off approach to managing complex coding tasks within Claude Code. It orchestrates a structured four-phase workflow—Target Definition, Planning, Execution, and Acceptance Review—using persistent state management. By monitoring quality gates (D1-D6) and status files, it intelligently transitions between phases, delegates to subagents, and applies adaptive retry logic. It is the ideal solution for developers who want to define a high-level objective and allow Claude to handle the iterative implementation, verification, and convergence process autonomously.

主な機能

01Four-phase automated workflow: Target, Planning, Execution, and Acceptance

02Persistent state management via .status.json and .target.md files

031 GitHub stars

04Intelligent subagent delegation for complex, multi-tiered task execution

05Seamless session recovery with cross-stage continuation and context rebuilding

06Automated D1-D6 quality gates with objective scoring and adaptive thresholds

ユースケース

01Automated requirement extraction and research for architectural changes

02Orchestrating multi-step developer workflows that require continuous planning and verification

03Hands-off end-to-end implementation of new features or complex bug fixes

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add huacheng/moonview auto

For use in Claude.ai and ChatGPT

Download Skill