yolo

Orchestrates unattended task execution via subagents with up-front questions and deferred unsafe actions.

1|Updated Jul 22, 2026
One-click install
npx skills add https://github.com/thititongumpun/skills --skill yolo-thititongumpun
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: yolo
Source: https://github.com/thititongumpun/skills/tree/main/skills/yolo
Command: npx skills add https://github.com/thititongumpun/skills --skill yolo-thititongumpun

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Running a long task while you are away normally stalls the moment the agent hits an ambiguity or needs confirmation. This Skill collects every blocking question in one up-front burst, then executes the entire plan-execute-review-fix loop unattended without ever waiting on a human again. ## Core Features & Use Cases - Single question burst: Scans the codebase first, then batches all scope, constraint, and approach questions into one AskUserQuestion call before going fully silent. - Subagent orchestration with model tiering: Follows the autopilot plan-execute-review-fix flow, assigning opus/fable planners, sonnet executors, and haiku for simple tasks based on Phase 1 complexity flags. - Safety by deferral: Unsafe or irreversible actions (force pushes, deploys, mass deletes, external writes) are skipped and logged with the exact command, never performed or worked around. - Use Case: Before leaving for the night, tell the agent to run a multi-file refactor unattended; you wake up to a summary of what was done, what was assumed, and which risky commands were deferred for your approval. ## Quick Start Tell the agent to run your task with yolo, for example: yolo migrate the API handlers to the new router while I'm asleep.

Frequently Asked Questions about yolo

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

FAQPage Schema
How do I run a Claude Code task unattended overnight?

Invoke the yolo skill with your task, answer the single batch of up-front questions it asks, then leave. It plans, executes via subagents, reviews, fixes, and finishes with a written summary without asking anything again.

What happens to dangerous commands like git push --force during an unattended run?

Unsafe or irreversible actions are never executed. The skill records the exact command and the reason it was deferred in a Deferred list, which appears in the final summary under what you need to do next.

Does yolo work outside Claude Code?

No. The skill is Claude Code only because it depends on subagents, AskUserQuestion, and the autopilot task file conventions that other harnesses do not provide.

How does yolo handle ambiguity when nobody is available to answer?

It resolves ambiguity using codebase conventions and its own judgment, then records each decision as an assumption in the final summary so you can review and reverse any call after you return.

Why did my unattended run use the wrong model for some tasks?

Model tiering is explicit: planners default to opus, executors to sonnet, with fable or haiku only when Phase 1 flags a task complex or simple. Leaving model unset is forbidden because an inherited session default silently retiers the run.