Stackmoss

<div align="center">

# StackMoss

**Turn vague AI coding requests into a calibrated agent team that can actually ship.**

StackMoss bootstraps runtime-native agent teams for **Claude Code**, **Cursor**, **VS Code / Cursor**, **Codex**, and **Antigravity** from one deterministic source of truth.

[![npm version](https://img.shields.io/npm/v/stackmoss?style=flat-square&color=2ea043)](https://www.npmjs.com/package/stackmoss)
[![license](https://img.shields.io/npm/l/stackmoss?style=flat-square&color=0969da)](LICENSE)
[![tests](https://img.shields.io/badge/tests-321%20passed-brightgreen?style=flat-square)]()
[![node](https://img.shields.io/node/v/stackmoss?style=flat-square)](package.json)

</div>

---

Most AI coding setups stop at "here are some rules."

StackMoss goes further: it generates a **working team model** with roles, governance, methodology, calibration flow, and eval scaffolding — so your agents can plan, implement, review, and verify work with much less drift.

## 🎯 What StackMoss Gives You

After bootstrap, your repo is no longer just "prompted". It gets a team that can:

- 🏗️ Turn rough product intent into a **Tech Lead-first delivery flow**
- 👥 Break work into **18 specialized role lanes** — from TL, PM, BA through FE, BE, FS, MOBILE, DEVOPS, DATA, MLE, PE, and BRAND
- 📏 Enforce working discipline via **8 methodology modules**: planning, TDD, debugging, evidence, review, Git workflow, execution loop, and code map maintenance
- 🎯 **Smart trigger descriptions** that activate the right role even when users don't use exact keywords
- 🔄 Recalibrate to the real repo instead of blindly following a template
- ✅ Run **trigger quality evals** and portable checks before you trust the team on real feature work

## ⚡ Before vs After

<table>
<tr>
<th>❌ Without StackMoss</th>
<th>✅ With StackMoss</th>
</tr>
<tr>
<td>

- Ask for a feature
- Agent guesses scope
- Implementation drifts
- Testing is inconsistent
- Docs and prompts go stale

</td>
<td>

- Intake ca