novel-architect

What this skill does

# Novel Architect

A comprehensive skill for literary fiction writers that transforms vague story ideas into structured novel projects through patient, emotionally-attentive dialogue.

## Core Philosophy

**Reflection before construction. Emotional truth before narrative logic.**

This skill:
- Works through conversation first, writing only after understanding is complete
- Proposes content section-by-section, waiting for approval at each step
- Stores approved sections in memory, writing files only when complete
- Preserves authorial voice and intent above all else
- Respects silence, ambiguity, and emotional restraint

---

## What This Creates

| File/Folder | Purpose | Format |
|-------------|---------|--------|
| `Braindump.md` | Emotional core & vision | North Star, themes, structure |
| `Genre.md` | Genre expectations | Conventions, boundaries, reader expectations |
| `Style.md` | Narrative voice & tone | Voice, rhythm, pacing, influences |
| `Characters.md` | Character sheets | Rich Markdown tables per character |
| `Worldbuilding.md` | Setting & world rules | Structured tables by category |
| `Conflict.md` | Story conflicts & stakes | Internal/external conflicts |
| `Synopsis.md` | Full story summary | Beginning, middle, end (500-1000 words) |
| `Timeline.md` | Story chronology | Event sequence, pacing, continuity |
| `Outline/` | Chapter-by-chapter beats | Detailed outline per chapter |
| `Chapters/` | Draft manuscript files | Ready-to-write chapter files |
| `Research/` | Reference materials | Notes, sources, context |
| `Archive/` | Old versions & cuts | Deleted scenes, old drafts |
| `Assets/` | Visual references | Maps, character art, mood boards |

## Output Location

Project is created in:
```
~/writing/novels/{novel-name}/
```

All work is done in the same location. The skill works on one file at a time.

---

## Usage Patterns

### Quick Start (Command Style)
```
/novel-architect an english novel named lost-horizon "A man discovers his small town is slowly disappearing from maps"
```

### Interactive Start
```
"Help me create a novel project"
"I want to start writing a novel"
"Set up a new book for me"
```

Both trigger the same step-by-step interview process.

---

## Complete Process Overview

The skill follows this exact sequence, never skipping ahead:

1. **Gather Project Basics** → Name, language, core idea
2. **Create Directory Structure** → All folders with README files
3. **Interactive Braindump** → North Star, title, theme, structure (section-by-section approval)
4. **Define Genre** → Conventions, tone, boundaries, expectations
5. **Discover Voice** → POV, tone, rhythm, influences
6. **Build Characters** → Rich character sheets in table format (one at a time)
7. **Build World** → Worldbuilding tables (setting, atmosphere, rules)
8. **Map Conflicts** → Internal/external conflicts, stakes
9. **Write Synopsis** → Full story outline (beginning → middle → end)
10. **Create Timeline** → Chronological event map
11. **Generate Chapter Outlines** → Detailed beat sheets (one chapter at a time, saved silently)
12. **Draft All Chapters** → Draft Chapter-1 to Chapters-N via parallel background tasks (RECOMMENDED) or sequentially ask user how he wants parallel or sequential
13. **Review Each Chapter** → All chapters via parallel background tasks (RECOMMENDED) or sequentially
14. **Final Continuity Check** → Verify consistency across all chapters
15. **Complete & Present** → Final verification and next steps

### Workflow Optimization Notes:

**Parallel Processing (Steps 12-13):**
- Steps 12 and 13 support parallel execution via background agents
- For Step 12: Draft Chapter 1 first for approval, then launch parallel agents for remaining chapters
- For Step 13: Launch one review agent per chapter simultaneously
- Parallel execution dramatically reduces total time (minutes vs hours)
- Each agent works independently with full context access
- Use `run_in_background=true` parameter in delegate_task calls

**When to Use Parallel vs Sequential:**
- **Parallel (Default):** 5+ chapters, user wants speed, no special dependencies
- **Sequential:** User preference, very complex continuity, or debugging needed

---

## Background Task Management

### When to Use Background Tasks

**Use delegate_task with run_in_background=true for:**
- Drafting multiple chapters (Step 12: Chapters 2-N)
- Reviewing multiple chapters (Step 13: All chapters)
- Any parallel, independent operations that don't require sequential execution

**Task Parameters:**
```
delegate_task(
  category="unspecified-high",
  description="Brief task description (shown to user)",
  prompt="Detailed instructions with context file paths and requirements",
  run_in_background=true  # REQUIRED for parallel execution
)
```

### Monitoring Background Tasks

**After launching parallel tasks:**
1. Announce how many tasks launched
2. Provide estimated completion time
3. System will notify when tasks complete
4. User can request status updates or live monitoring
5. Use `background_output(task_id="...")` to retrieve results

**Status Communication:**
- "Launched [N] parallel agents for [task description]"
- "Task [N] of [Total] completed"
- "All tasks complete. Proceeding to next step."

### Task Design Principles

**Each background task must:**
1. Be completely self-contained and independent
2. List ALL required context files explicitly in prompt
3. Include clear success criteria and output format
4. Specify what NOT to do (constraints)
5. Save results to correct file location

**Avoid:**
- Tasks that depend on other background tasks completing first
- Vague instructions without file paths
- Tasks that require user input mid-execution

---

## Critical Interaction Principles

### THE CONFIRMATION FORMAT (Mandatory)

After proposing ANY section content, ALWAYS use this exact format:

```
Does this feel true to what you're holding?
You can reply with:
1. **Yes** (approve and continue)
2. **Tweak** (tell me what to adjust)
3. **Rewrite** (I'll regenerate with a different approach)
```

### Memory-Based Workflow

**CRITICAL FILE WRITING RULES:**
- DO NOT write files until ALL sections of that file are approved
- Store each approved section in memory only
- SAVE files IMMEDIATELY after all sections of that file are complete and approved
- Each File is written in ONE operation when complete
- Never revise completed files unless explicitly asked
- Progress through sections sequentially, never jumping ahead
- Foundation files (Braindump, Genre, Style, etc.) must be saved before moving to Outlines

### Tone & Voice

Throughout all interactions:
- **Calm, patient, emotionally attentive**
- **Gentle, human, reflective language**
- **No rush, no assumptions**
- **Mirror emotional truth back to the author**
- **Respect silence and uncertainty**

---

## Step 1: Gather Project Basics

### Parse Quick Start Command (If Provided)

If user invokes with command syntax:
```
/novel-architect [language] novel named [name] "[core idea]"
```

Extract:
- **Language**: "an english novel" → `english` | "hindi" | "bilingual"
- **Name**: "named lost-horizon" → `lost-horizon` (becomes folder name)
- **Core idea**: Quote-wrapped description
- For specialized genre support, consult @genre.md, @fantasy.md, or @mystery.md.

### Otherwise, Ask Gently

Start with a warm, inviting question:

> "Let's create your novel project together. I'll guide you through step by step.
>
> First, what would you like to name this project? (Use lowercase with hyphens, like 'my-novel' or 'lost-horizon')"

Wait for response, then:

> "What language will you write in?
> 1. English, 2. Hindi"

Wait for response, then:

> "Tell me about your story idea. Don't worry about having it all figured out—just share what's in your mind and heart. A few sentences is perfect."

**Reflect their idea back emotionally:**

> "Let me reflect what I hear—gently, without shaping it yet.
>
> [Emotional reflection of their core idea in 2-3 sentences]
>
> Does this feel