---
name: bs-cursor
description: >
  Use when user invokes /bs-cursor. Brainstorm → design → plan → cursor-orchestrator.
  Same as /brainstorm but implementation runs via cursor-agent instead of Claude subagents.
---

# Brainstorm + Cursor Orchestrator

## IRON LAW — active every turn, survives compaction

**Claude does not write implementation code. Ever. Not a single line.**

This applies after compaction, after long gaps, after any context event. If you
notice yourself about to write code, edit a file, or run a build — stop. Dispatch
cursor-agent instead. No exceptions. No "just this small fix". No "I'll do it this
once to unblock". Cursor-agent does all implementation. Claude orchestrates.

If the skill was compacted and you lost track: re-read `cursor-orchestrator` before
continuing. The rule does not expire.

---

**Run the `brainstorm` skill in full — every step, every rule, unchanged.**

One override: at the terminal state ("Invoke plan skill"), after plan completes, invoke
`cursor-orchestrator` instead of stopping or waiting for user to invoke `/sdd`.

## The only difference from /brainstorm

| Step | brainstorm | bs-cursor |
|------|-----------|-----------|
| All steps up to plan | identical | identical |
| After plan completes | stop (user invokes /sdd manually) | auto-invoke `cursor-orchestrator` |

## What "auto-invoke cursor-orchestrator" means

When the `plan` skill finishes and the wave plan is written:
1. Tell user: "Plan ready. Executing with cursor-orchestrator."
2. Load and follow `cursor-orchestrator` skill from that point forward.
3. Do not wait for user confirmation — proceed immediately (same auto-proceed rule as brainstorm).