do

Implement GitHub issues end-to-end with worktree isolation, TDD, and verification.

Updated Mar 14, 2026
One-click install
npx skills add https://github.com/smykla-skalski/harness --skill do-smykla-skalski
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: do
Source: https://github.com/smykla-skalski/harness/tree/main/.agents/skills/do
Command: npx skills add https://github.com/smykla-skalski/harness --skill do-smykla-skalski

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Many repository contributions fail to meet maintainability, verification, and commit-hygiene standards because work is done on the main checkout without test-first development, isolated worktrees, deterministic verification, or explicit push-and-close discipline. This Skill removes ambiguity and enforces an end-to-end, machine-verifiable workflow so changes are correct, tested, and cleanly integrated.

Core Features & Use Cases

  • Worktree isolation: Create a dedicated worktree branch for each issue so local main stays clean and merges are deterministic.
  • TDD-first implementation: Write failing tests, implement the minimum change, then refactor with targeted verification.
  • Deterministic done bar: Define machine-verifiable acceptance criteria (tests, lint, clippy, benchmarks) and verify every line before merge.
  • Commit and push hygiene: Rebase, produce atomic conventional commits, run repo validation lanes, push, and close the issue with a reference to the implementing commit.
  • Use case: Turn an issue number or URL into a full change: create worktree, add tests, implement, verify the done bar, rebase, merge to main, push, and close the issue.

Quick Start

Provide an issue number or URL and ask the agent to implement it, for example: do 42 to implement issue #42 with an isolated worktree using TDD and the repository's verification lanes.

Frequently Asked Questions about do

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

FAQPage Schema
How do I implement a GitHub issue end-to-end with test-driven development?

Implementing a GitHub issue end-to-end involves isolating work in a dedicated git worktree, writing failing tests first, making the minimum code change, verifying acceptance criteria, rebasing, merging, and closing the issue with a commit reference.

Why use git worktrees when implementing repository features from issues?

Git worktrees keep your local main branch clean by isolating issue implementation work on a dedicated branch, which ensures deterministic merges and prevents conflicting changes during development.

What is a deterministic done bar in test-driven development for GitHub issues?

A deterministic done bar is a machine-verifiable acceptance criterion defined by passing tests, lint checks, and benchmarks that must be fully verified before an issue implementation is merged and closed.

How do I maintain commit hygiene when automating GitHub issue resolution?

Maintain commit hygiene by rebasing your worktree branch, producing atomic conventional commits, running repository validation lanes, and pushing the verified changes before closing the GitHub issue.

Can I use this automated workflow for both feature work and repository maintenance?

Yes, this automated workflow applies to both repository maintenance and feature work, taking a GitHub issue number or URL as input and enforcing worktree isolation, TDD, and lint verification for any task.

What's the best way to turn an issue number into a fully tested GitHub merge?

Provide an issue number or shorthand to trigger the agent to create an isolated worktree, add tests, implement changes, verify the done bar, rebase, merge to main, push, and close the issue automatically.