Large SYSTEM_PROMPT causing context overflow

## Problem Description

The copilot system is experiencing context overflow issues due to an extremely large `SYSTEM_PROMPT` that gets added to every chat conversation. This is causing:

- **Context limit exhaustion** - The prompt fills up available context quickly
- **Performance degradation** - Slower response times
- **Higher API costs** - More tokens per request
- **Poor user experience** - Copilot becomes unresponsive

## Root Cause Analysis

The `SYSTEM_PROMPT` in `/apps/rowboat/src/application/lib/copilot/copilot.ts` includes:

1. **`COPILOT_INSTRUCTIONS_MULTI_AGENT`** (263 lines) - ✅ Acceptable
2. **`COPILOT_MULTI_AGENT_EXAMPLE_1`** (1,329 lines) - ❌ **MAJOR ISSUE**
3. **`CURRENT_WORKFLOW_PROMPT`** (13 lines) - ✅ Acceptable
4. **Auto-loaded docs** from `copilot_multi_agent.ts` - ❌ **POTENTIAL ISSUE**

**Total context size:** ~1,600+ lines added to every chat!
## Critical Issue: Token Consumption

**Current Problem:** With a simple chat, we're hitting the **30,000 token limit** immediately due to the massive `SYSTEM_PROMPT`. This means:

- Users can't have meaningful conversations
- Copilot becomes unresponsive after just a few messages
- The system is essentially unusable for any real workflow

**Token Breakdown:**
- SYSTEM_PROMPT: ~25,000+ tokens (1,600+ lines)
- User message: ~100-500 tokens
- **Total: 30,000+ tokens** (hitting the limit!)

Start by reading /apps/rowboat/src/application/lib/copilot/copilot.ts and tracing how COPILOT_INSTRUCTIONS_MULTI_AGENT, COPILOT_MULTI_AGENT_EXAMPLE_1, and CURRENT_WORKFLOW_PROMPT are assembled. Then inspect copilot_multi_agent.ts for auto-loaded documentation and measure the resulting prompt size. Done should mean ordinary chats no longer exhaust the 30,000-token context limit while retaining the required copilot behavior.

Written by the indexing model from the issue text.