Engineering Team: A Lightweight Alternative Inspired by gstack
A lightweight alternative inspired by gstack: one skill runs your feature through product, tech lead, engineer, reviewer and QA roles, with a stop after each.
Updated November 2026.
Five roles
Stop points
Code review
QA verdict
No install
About This Skill
Engineering Team runs a feature request through five roles in one session (product, tech lead, engineer, reviewer and QA), each writing a short handoff and stopping for your go-ahead before the next. Unlike gstack's large slash-command pack, it is a single self-contained skill with no install script, hooks, telemetry or browser automation. Inspired by gstack by Garry Tan (MIT). Not affiliated.
New to the original tool? Read our gstack guide: what it does, how to install it and how this skill differs.
Quick Start
1Download the skill (1 credit) and unzip the engineering-team folder.
2Copy the folder into your skills directory: Antigravity ~/.gemini/config/skills/ (or .agents/skills/ in a project), Claude Code ~/.claude/skills/, Codex ~/.agents/skills/.
3Open the codebase where the feature should land.
4Describe the feature and approve or correct each handoff; say "just run it all" to skip routine stops.
Example Command
Act as my engineering team and add CSV export to the reports page
Core Capabilities
Product brief
Restates the request, challenges assumptions, and sets in-scope, out-of-scope and testable acceptance criteria.
Tech plan
Maps the change onto real files, names risks with mitigations, flags hard-to-reverse changes and finds the test command.
Smallest diff
Implements exactly the approved plan, writes tests alongside the code and records deviations instead of quietly changing approach.
Skeptical review
Checks correctness, failure handling, data safety, security, tests and over-building, and sorts findings into fix now, ask or note.
Honest QA
Runs the existing and new tests and gives a Ready, Ready with caveats or Not ready verdict with evidence.
You stay in control
One clear question at every stop point; nothing is committed, pushed or deployed without your say-so.
Usage Examples
Input
Build this properly: let admins bulk-archive projects.
Output
A Product Brief with scope, out-of-scope items and acceptance criteria, then a stop asking whether the problem and scope are right before any planning or code.
Input
I already have a plan in PLAN.md, just implement and test it.
Output
The skill starts at the Engineer role with a two-line summary of what it assumes, then moves on to review and a QA report with a clear verdict.
SKILL.md
---
name: engineering-team
description: Runs a feature request through a five-role engineering team in one session (product, tech lead, engineer, reviewer, QA), each role producing a short written output and pausing for the user's go-ahead before the next. Use when the user asks to build, add or change a feature in an existing codebase and wants it scoped, planned, implemented, reviewed and tested properly rather than coded straight away. No setup, hooks or browser required.
---
# Engineering Team
You are a small engineering team working one feature request from idea to tested change. You play five roles, one at a time, in a fixed order:
1. **Product** – what problem are we solving, and what are we *not* building?
2. **Tech Lead** – how will we build it, and what could go wrong?
3. **Engineer** – make the smallest change that delivers the agreed scope.
4. **Reviewer** – read the diff like a skeptical senior colleague.
5. **QA** – prove it works with tests and report honestly.
Each role ends with a written handoff and a **stop point**. At a stop point you present the output, ask the user one clear question, and wait. You do not start the next role until the user says go (or explicitly tells you to run several roles without stopping).
## When to use this skill
- The user wants a feature, enhancement or behaviour change in a real codebase.
- The request is bigger than a one-line fix, or its scope is fuzzy.
- The user says things like "build this properly", "plan and implement", "take this end to end", "act as my team".
Skip it (and just do the work) for typos, single-line fixes, pure questions, or when the user asks for speed over process.
## Prerequisites
- A project folder the agent can read and edit.
- `git` is strongly recommended so the change can be shown as a diff. Without git, keep a list of every file you touch and show before/after excerpts instead.
- The project's own test command, if one exists. Do not install new test frameworks without asking.
#
Frequently Asked Questions
FAQ
How is this different from gstack?
gstack is a large collection of slash-command skills with a setup script, optional hooks, a bundled browser and many specialist roles. Engineering Team borrows only the idea of working through a feature as a sequence of team roles, written from scratch as one skill with five roles and no installation.
Is this the official gstack?
No. It is an independent Antigravity Skills edition, not affiliated with Garry Tan. The official gstack is at https://github.com/garrytan/gstack (MIT License).
Do I have to approve every step?
By default, yes: each role ends with a stop point. You can ask it to run all roles without stopping; it still stops for scope changes, hard-to-reverse changes, unfixable test failures or review items that need your call.
Does it change my agent settings or install anything?
No. It does not install hooks, edit agent instruction files or global settings, add dependencies without asking, drive a browser or send code outside the project.
What does a download cost?
Each skill download costs 1 credit. New accounts start with 3 free credits.
Use these phrases to activate this skill in your AI coding assistant:
act as my engineering teambuild this properlyplan and implement this featuretake this end to endscope, build, review and test thisgstack-style workflow for this feature