Shall We? – A pre-engineering sanity check for AI-era feature bloat

███████╗ ██╗  ██╗  █████╗  ██╗      ██╗         ██╗    ██╗ ███████╗ ██████╗ 
██╔════╝ ██║  ██║ ██╔══██╗ ██║      ██║         ██║    ██║ ██╔════╝ ╚════██╗
███████╗ ███████║ ███████║ ██║      ██║         ██║ █╗ ██║ █████╗       ██╔╝
╚════██║ ██╔══██║ ██╔══██║ ██║      ██║         ██║███╗██║ ██╔══╝      ╚═╝  
███████║ ██║  ██║ ██║  ██║ ███████╗ ███████╗    ╚███╔███╔╝ ███████╗    ██╗  
╚══════╝ ╚═╝  ╚═╝ ╚═╝  ╚═╝ ╚══════╝ ╚══════╝     ╚══╝╚══╝  ╚══════╝    ╚═╝

A skill that helps you decide whether an idea is worth building — before you build it.

Implements the Agent Skills open standard. Works with Claude Code, Cursor, GitHub Copilot, Gemini CLI, OpenAI Codex, and any other compatible agent.

We're in an AI hype cycle where everyone wants to build everything, just because they can. Features get shipped because they're technically possible. Products get started because the demo looked good. Nobody stops to ask the uncomfortable question.

Shall we, though?

This skill is the uncomfortable question. It was inspired by two characters who've been fighting this battle from different angles:

• The caveman(grugbrain.dev) — who's been sayingnoto complexity since before it was cool, and whose club has saved more codebases than any design pattern - The ponytail— the architect who's seen enough hype cycles to know that most ideas aren't bad, they're just badly timed, badly scoped, or solving the wrong problem

shall-we

thinks like both of them, but operates one level higher: at the PM layer, where the decision to build (or not) is still cheap. It asks the questions that should happen before any code is written, any ticket is created, any design doc is started.

• Restates your idea to confirm it understood correctly
• Asks one question at a time — only what it needs to know
• Stops early if a deal-breaker surfaces
• Delivers a structured verdict:

## Verdict: YES / NO / LATER

**Why:** ...

**Strengths:**
- ...

**Concerns:**
- ...

**What would need to be true for a YES:**
- ...

The four lenses: business value, market demand, complexity, and risk.

/plugin marketplace add Tacsiazuma/shall-we
/plugin install shall-we@shall-we

Then invoke with /shall-we

.

gemini skills install https://github.com/Tacsiazuma/shall-we.git --consent

Clone the skill into Cursor's skill directory:

git clone https://github.com/Tacsiazuma/shall-we.git /tmp/shall-we
cp -r /tmp/shall-we/skills/shall-we ~/.cursor/skills/shall-we

Cursor picks it up automatically — no restart needed.

Copy the skill into your project:

mkdir -p .agents/skills
git clone https://github.com/Tacsiazuma/shall-we.git /tmp/shall-we
cp -r /tmp/shall-we/skills/shall-we .agents/skills/shall-we
mkdir -p ~/.agents/skills
git clone https://github.com/Tacsiazuma/shall-we.git /tmp/shall-we
cp -r /tmp/shall-we/skills/shall-we ~/.agents/skills/shall-we

The skill lives at skills/shall-we/SKILL.md

in this repo. Copy or symlink that directory wherever your agent looks for skills. See agentskills.io/clients for agent-specific paths.

.
├── skills/
│   └── shall-we/
│       └── SKILL.md       # agentskills.io standard skill
├── .claude-plugin/
│   └── marketplace.json   # Claude Code marketplace manifest
└── plugins/
    └── shall-we/
        └── .claude-plugin/
            └── plugin.json