solo

Guides single-agent task execution through dialogue, planning, implementation, and user-confirmed completion.

Updated Nov 21, 2025
One-click install
npx skills add https://github.com/alatyshau/duet --skill solo-alatyshau
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: solo
Source: https://github.com/alatyshau/duet/tree/main/packages/skills/workflows/solo
Command: npx skills add https://github.com/alatyshau/duet --skill solo-alatyshau

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When working on simple tasks with an AI assistant, there is often no clear structure for when to plan, when to just execute, and when a task is truly finished. This Skill defines a lightweight single-agent workflow that keeps work organized without the overhead of multi-agent coordination. ## Core Features & Use Cases - Structured Flow: Follows a clear cycle of DIALOGUE, optional PLANNING, EXECUTE, report, and DONE phases tied to topic files and spec/ documentation. - Planning Decision Rules: Defines explicit criteria for when planning can be skipped (single-file changes, clear specs, obvious bug fixes) versus when it is required (multi-file changes, architecture decisions). - Human-Gated Completion: Enforces the rule that only the user can mark a task as DONE, preventing premature closure by the agent. - Use Case: You ask the assistant to fix a small bug in one file. The Skill skips planning, implements the fix with tests, updates spec/ if needed, and waits for your explicit approval before closing the task. ## Quick Start Ask the assistant to work on a simple task using the solo workflow and confirm completion only after you review the result.

Frequently Asked Questions about solo

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

FAQPage Schema
How do I structure a simple AI-assisted coding task?

Use a solo workflow with phases: dialogue to clarify requirements, optional planning, execution with code and tests, then a report to the user. The task closes only after explicit user confirmation, keeping a clear audit trail in topic files.

When should I skip planning before implementing a change?

Skip planning for single-file changes, tasks with exact user-provided specs, or bug fixes with obvious solutions. Planning is required when multiple files are affected, architecture decisions are needed, or the user explicitly asks to plan first.

Can an AI agent mark a task as done by itself?

No. In this workflow the agent never marks a step as DONE. Only the human can close a task by explicitly saying it is done or approved, ensuring the user reviews all work before completion.

When should I switch from a solo workflow to a multi-agent workflow?

Switch to a pair or multi-agent workflow like SDDG when changes are complex or risky and need an external AI reviewer, or when multiple perspectives on the solution are wanted.

How do specs stay in sync with code changes?

The spec/ directory is the source of truth and must be updated whenever architecture changes. Each commit should contain code and spec updates together so they remain in integrity, while the topic file is treated as temporary.