skills.rest
arndvs
arndvsCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

do-work

Implements tasks through a plan, test, validate, commit, and preflight execution loop.

1|Updated Mar 26, 2026
Search Tags:#tdd#pull request#validation#implementation#git commits#coding agent#execution loop
One-click install
npx skills add https://github.com/arndvs/ctrlshft --skill do-work-arndvs
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: do-work
Source: https://github.com/arndvs/ctrlshft/tree/main/skills/do-work
Command: npx skills add https://github.com/arndvs/ctrlshft --skill do-work-arndvs

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Coding agents often drift mid-task, skip validation, or produce unreviewed commits. This Skill provides a structured execution loop that takes a task from understanding through implementation, validation, atomic commits, and pre-PR preflight checks. ## Core Features & Use Cases - Structured execution pipeline: Understand the task, plan in vertical slices, implement with red/green/refactor TDD for backend code, and validate using auto-detected project feedback loops (npm, composer, make, pytest). - Commit and ship workflow: Commits via the atomic-commits skill and runs the pr-preflight skill before pushing and opening a PR in Ship mode. - Context handoff: When context usage exceeds 40% or a phase ends, persists the remaining plan to working/active/ so a fresh session can resume. - Use Case: Given a GitHub issue to implement, the agent reads the issue, writes one failing test at a time, runs the project's test and lint scripts, commits atomically, and opens a reviewed PR. ## Quick Start Ask the agent to do work on the current GitHub issue and follow the implement, validate, and commit loop until a PR is opened.

Frequently Asked Questions about do-work

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

FAQPage Schema
How do I implement a GitHub issue with an AI coding agent?▼

Invoke the do-work loop with the issue reference. The agent reads the issue, plans vertical slices, implements with one failing test at a time for backend code, runs the project's validation scripts, and commits atomically before opening a PR.

How does the skill detect which test and lint commands to run?▼

It auto-detects feedback loops from workspace files: package.json scripts run with the project's package manager, composer.json scripts via composer run, Makefile targets via make, and pytest, mypy, or ruff for Python projects.

Does the do-work skill use TDD for all code?▼

TDD with red/green/refactor applies to backend code only, writing one failing test at a time in tracer-bullet style. Frontend code is implemented directly without TDD.

What happens when context usage gets too high during a task?▼

When context exceeds 40% or a phase completes, the agent commits all work, persists the remaining plan to working/active/, and provides a pickup command so a fresh session can resume exactly where it stopped.

When should I skip the pr-preflight step?▼

Skip preflight when using Commit mode for intermediate checkpoints. Preflight runs once on the final slice before pushing, auditing the full main..HEAD diff so review happens in one pass.

Related Skills

Explore More Software Engineering Skills→
zeshonezeshone

prd-to-issues

Convert PRDs into dependency-ordered GitHub issues with vertical slices.

Community
Advanced
JacobPEvansJacobPEvans
3

rebase-pr

Rebase pull requests onto main with signed commits and GitHub API checks.

Community
Advanced
atelier-fashionatelier-fashion
159

reflect

Audit recent code changes for correctness, conventions, architecture, and testing.

Official
Advanced
EmrebenerEmrebener

ss-agile-implementing-an-issue

Implement a GitHub issue on a checked-out feature branch with commits and verification.

Community
Advanced