bookstrap-plan-write

name	bookstrap-plan-write
description	Generate writing tasks from BRD structure and corpus, creating chapter/section tasks with dependencies, pre-write queries, and consistency checks
disable-model-invocation	true
allowed-tools	Bash, Read

/bookstrap-plan-write - Generate Writing Tasks

Analyze the Book Requirements Document (BRD) structure and existing corpus to generate prioritized writing tasks with full metadata, dependencies, and consistency checks.

Purpose

Transform the BRD outline into executable writing tasks, each specifying required knowledge, pre-write database queries, post-write storage operations, and consistency constraints. Tasks are generated with dependencies (blockedBy relationships) to ensure proper writing order.

This command delegates the planning work to the outline-planner agent.

Input Arguments

None. This command reads from the database:

• BRD structure and chapter outline
• Existing corpus and coverage
• Entities and relationships already extracted
• Resolved knowledge gaps

Processing Workflow

1. Load BRD Structure

Query the database to retrieve the Book Requirements Document:

# Query BRD from database
surreal sql --conn http://localhost:2665 \
  --user root --pass root \
  --ns bookstrap --db <database-name> \
  --query "SELECT * FROM brd ORDER BY version DESC LIMIT 1;"

Extract structural requirements:

• Planned chapters and sections
• Chapter sequence and dependencies
• Section topics and themes
• Target word counts per section
• Voice, tone, and style requirements
• POV and tense specifications

2. Analyze Corpus Coverage

Check what knowledge exists to support each planned section:

-- Coverage by planned chapter/topic
SELECT chapter, count() as sources
FROM section->cites->source
GROUP BY chapter;

-- Entity availability
SELECT * FROM character;
SELECT * FROM location;
SELECT * FROM event ORDER BY sequence;
SELECT * FROM concept;

-- Resolved knowledge gaps
SELECT * FROM knowledge_gap
WHERE resolved = true;

-- Source distribution
SELECT source_type, reliability, count() as total
FROM source
GROUP BY source_type, reliability;

-- Timeline completeness
SELECT sequence, count() as events
FROM event
GROUP BY sequence
ORDER BY sequence;

3. Generate Chapter/Section Writing Tasks

For each chapter/section in the BRD, create a writing task with comprehensive metadata:

Task Metadata Structure

{
  "taskId": "chapter-3-scene-2",
  "subject": "Write Chapter 3, Scene 2: The Confrontation",
  "description": "Write the confrontation scene between Anna and Erik at the safehouse, revealing the betrayal and establishing the stakes for the final act.",
  "blockedBy": ["chapter-3-scene-1"],
  "metadata": {
    "chapter": 3,
    "section": 2,
    "sequence": 12,
    "target_words": 2500,
    "scene_type": "dialogue_heavy",
    "required_knowledge": [
      "character:anna.background",
      "character:anna.status",
      "character:erik.background",
      "character:erik.relationships",
      "location:safehouse.description",
      "location:safehouse.introduced",
      "event:discovery.details",
      "event:discovery.sequence",
      "concept:betrayal_theme"
    ],
    "pre_queries": [
      "SELECT * FROM section WHERE embedding <|5|> $theme_vector ORDER BY vector::similarity(embedding, $theme_vector) DESC",
      "SELECT * FROM character:anna->knows->character",
      "SELECT * FROM character:erik->knows->character",
      "SELECT * FROM event WHERE sequence < 12 ORDER BY sequence DESC LIMIT 5",
      "SELECT * FROM section WHERE chapter = 3 AND sequence < 2 ORDER BY sequence"
    ],
    "post_writes": [
      "CREATE section SET content=$content, embedding=$vec, chapter=3, sequence=12, word_count=$word_count, status='draft'",
      "RELATE character:anna->confronts->character:erik SET context=$confrontation_context",
      "RELATE event:confrontation->follows->event:discovery SET narrative_link=true",
      "RELATE section:$section_id->cites->source:$source_ids",
      "RELATE section:$section_id->appears_in<-character:anna",
      "RELATE section:$section_id->appears_in<-character:erik",
      "RELATE section:$section_id->located_in->location:safehouse"
    ],
    "consistency_checks": [
      "character:anna.status != 'dead'",
      "character:erik.status != 'dead'",
      "location:safehouse.introduced = true",
      "event:discovery.sequence < 12",
      "character:anna->knows->character:erik EXISTS"
    ],
    "voice_requirements": {
      "pov": "third_limited",
      "tense": "past",
      "tone": "tense_suspenseful",
      "comparable": "tension like le Carré"
    }
  }
}

4. Determine Task Dependencies

Establish blockedBy relationships based on:

Sequential Dependencies:

• Chapter N blocks Chapter N+1
• Section M blocks Section M+1 within same chapter
• Chronological event order (flashbacks excepted)

Knowledge Dependencies:

• Character introduction blocks appearances
• Location introduction blocks scenes set there
• Event establishment blocks references to it
• Concept definition blocks thematic callbacks

Graph Dependencies:

• Relationship creation blocks relationship references
• Timeline placement blocks temporal references

5. Specify Required Knowledge

For each task, list the entities and facts that MUST exist in the database before writing:

"required_knowledge": [
  "character:protagonist_id.name",
  "character:protagonist_id.description",
  "character:protagonist_id.background",
  "character:protagonist_id.status",
  "location:setting_id.description",
  "location:setting_id.introduced",
  "event:prior_event_id.sequence",
  "concept:theme_id.description"
]

If any required knowledge is missing, the task should be flagged as blocked by a knowledge gap.

6. Generate Pre-Write Queries

For each task, specify the database queries to run BEFORE writing to gather context:

Semantic Search Queries:

-- Find thematically similar passages
SELECT * FROM section
WHERE embedding <|5|> $theme_vector
ORDER BY vector::similarity(embedding, $theme_vector) DESC;

-- Find related concepts
SELECT * FROM concept
WHERE embedding <|3|> $concept_vector;

Graph Traversal Queries:

-- Character relationships
SELECT * FROM character:$char_id->knows->character;
SELECT * FROM character:$char_id<-appears_in<-section ORDER BY sequence;

-- Source support
SELECT * FROM source WHERE ->supports->concept:$concept_id;

-- Location context
SELECT * FROM location:$loc_id<-located_in<-section;

Timeline Queries:

-- Prior events
SELECT * FROM event
WHERE sequence < $current_sequence
ORDER BY sequence DESC
LIMIT 5;

-- Chronological section order
SELECT * FROM section
WHERE chapter = $chapter AND sequence < $current_sequence
ORDER BY sequence;

7. Specify Post-Write Operations

For each task, define the database operations to execute AFTER writing:

Content Storage:

CREATE section SET
  content = $content,
  embedding = $embedding_vector,
  chapter = $chapter_num,
  sequence = $section_num,
  word_count = $word_count,
  status = 'draft',
  created_at = time::now();

Entity Extraction and Creation:

-- Extract and create new entities mentioned in the section
-- (Performed by writer agent using LLM entity extraction)

-- Link existing entities
RELATE section:$section_id->appears_in<-character:$char_ids;
RELATE section:$section_id->located_in->location:$loc_ids;

Relationship Creation:

-- Character interactions
RELATE character:$char1->interacts_with->character:$char2
  SET context = $interaction_description;

-- Event sequencing
RELATE event:$event_id->follows->event:$prior_event_id;

Source Citations:

-- Link to supporting sources
RELATE section:$section_id->cites->source:$source_ids
  SET claim = $cited_claim;

Timeline Updates:

-- Update event sequence
UPDATE event:$event_id SET sequence = $sequence_num;

8. Define Consistency Checks

For each task, specify constraints that MUST be true before writing:

Entity State Checks:

character:$char_id.status != 'dead'
character:$char_id.introduced = true
location:$loc_id.introduced = true

Relationship Checks:

character:$char1->knows->character:$char2 EXISTS
character:$char_id->located_at->location:$loc_id EXISTS

Timeline Checks:

event:$prior_event.sequence < $current_sequence
event:$event_id.date < time::now()

Contradiction Checks:

-- No conflicting facts
SELECT * FROM section
WHERE ->cites->source->contradicts->source<-cites<-section:$section_id;

9. Store Tasks in Database

Create task records with all metadata:

# Tasks are stored in Harness task system via TaskCreate
# Metadata is stored as JSON in the task's metadata field

10. Delegate to Outline Planner

Invoke the outline-planner agent to perform the detailed planning:

# Load outline-planner agent with context
# Agent will:
# 1. Read BRD structure from database
# 2. Execute corpus coverage queries
# 3. Generate chapter/section tasks
# 4. Determine dependencies and blockedBy
# 5. Specify required knowledge per task
# 6. Generate pre-write queries
# 7. Define post-write operations
# 8. Set consistency checks
# 9. Store tasks via TaskCreate
# 10. Generate report

The outline-planner agent has read-only database access and uses the outlining and surrealdb skills.

Output Format

Report generated writing tasks to the user:

WRITING PLAN GENERATED
======================

BRD: <Book Title>
Structure: <N> chapters, <M> sections
Total Word Target: <X> words

CHAPTER BREAKDOWN
-----------------
Chapter 1: Introduction
  Section 1: Opening (Task #1) - 1,500 words
    Required: character:anna, location:london
    Blocked by: None

  Section 2: Training Begins (Task #2) - 2,000 words
    Required: character:anna, location:beaulieu, event:recruitment
    Blocked by: Task #1

Chapter 2: Deployment
  Section 1: The Flight (Task #3) - 1,800 words
    Required: character:anna, location:france, event:deployment
    Blocked by: Task #2

  Section 2: First Contact (Task #4) - 2,200 words
    Required: character:anna, character:henri, location:safehouse
    Blocked by: Task #3

[... continued for all chapters ...]

TASK SUMMARY
------------
Total Tasks: 32
Ready to Write: 1 (Task #1)
Blocked by Dependencies: 31
Blocked by Knowledge Gaps: 0

DEPENDENCIES GRAPH
------------------
Task #1 → Task #2 → Task #3 → Task #4 → ...
       ↘ Task #5 → Task #6 → ...

KNOWLEDGE REQUIREMENTS
----------------------
All required knowledge present in corpus:
  ✓ 23 characters defined
  ✓ 45 locations described
  ✓ 88 events sequenced
  ✓ 34 concepts documented

NEXT STEPS
----------
Run /bookstrap-write to begin autonomous writing.
Tasks will execute in dependency order.
Use /bookstrap-status to monitor progress.

Statistics to Track

Calculate and report:

• Total chapters and sections planned
• Total target word count
• Tasks ready to write (no blockedBy)
• Tasks blocked by dependencies
• Tasks blocked by knowledge gaps
• Required entities and their availability
• Estimated writing time (sections × avg time)
• Coverage percentage (corpus vs. requirements)

Implementation Notes

Agent Delegation

This command is a thin wrapper that:

1. Verifies database connection
2. Checks that BRD exists
3. Verifies sufficient corpus coverage
4. Invokes the outline-planner agent
5. Displays the agent's output

The actual planning logic lives in the outline-planner agent to keep concerns separated.

Task Generation Strategies

The outline-planner uses multiple approaches:

• BRD structure parsing: Extract chapter/section outline
• Corpus analysis: Determine what knowledge supports each section
• Dependency mapping: Build task graph from sequential/knowledge dependencies
• Query generation: Create semantic/graph/timeline queries per task
• Metadata enrichment: Add voice, tone, style requirements from BRD

Idempotency

Re-running /bookstrap-plan-write should:

• Update existing tasks if BRD changed
• Add new tasks if structure expanded
• Not duplicate existing tasks
• Re-evaluate dependencies based on current corpus state
• Update blockedBy if knowledge gaps resolved

Configuration

Writing task generation can be configured in bookstrap.config.json:

{
  "writing": {
    "default_section_words": 2000,
    "min_sources_per_section": 2,
    "require_knowledge_checks": true,
    "auto_generate_queries": true,
    "strict_dependencies": true
  }
}

Error Handling

Error	Recovery
BRD not found	Abort with message to run /bookstrap-init first
BRD missing structure	Abort with message that BRD needs chapter outline
Database connection failed	Abort with message to start SurrealDB
Insufficient corpus	Warning and flag tasks as blocked by knowledge gaps
Outline-planner agent error	Display error, suggest re-running or checking logs

Pre-requisites

Before running /bookstrap-plan-write:

1. BRD created: /bookstrap-init must have been run with structure defined
2. SurrealDB running: Database must be accessible
3. Schema initialized: Database schema loaded
4. Corpus sufficient: /bookstrap-ingest and /bookstrap-research completed
5. Knowledge gaps resolved: All high-priority gaps from /bookstrap-plan-research addressed

Related Commands

• /bookstrap-init - Create BRD with chapter structure
• /bookstrap-ingest - Load initial corpus
• /bookstrap-plan-research - Identify and fill knowledge gaps
• /bookstrap-research - Execute research tasks to fill gaps
• /bookstrap-write - Execute writing tasks generated by this command
• /bookstrap-status - View writing progress

Supporting Agents

Agent	Role
outline-planner	Generates writing tasks with full metadata and dependencies

Supporting Skills

Skill	Purpose
outlining/	Story structure, planning strategies, task generation
surrealdb/	Database query patterns for coverage analysis

Example Usage

# After research complete, generate writing plan
/bookstrap-plan-write

# Re-run after BRD structure changes
/bookstrap-init  # Update BRD outline
/bookstrap-plan-write  # Regenerate tasks

# Check plan before starting to write
/bookstrap-plan-write
/bookstrap-status

Integration with Writing Loop

This command bridges research and writing:

init → ingest → plan-research → research → plan-write → write → edit
                     ↑                          │            │
                     └────── gaps found ────────┴────────────┘

Once tasks are generated, /bookstrap-write executes them autonomously in dependency order. If new knowledge gaps are discovered during writing, the writer flags them, and the flow returns to research planning.

Task Metadata Reference

Complete metadata structure for each writing task:

{
  "taskId": "unique-task-id",
  "subject": "Brief task description",
  "description": "Detailed task description with context",
  "activeForm": "Writing Chapter X, Section Y",
  "blockedBy": ["task-id-1", "task-id-2"],
  "metadata": {
    "chapter": 1,
    "section": 1,
    "sequence": 1,
    "target_words": 2000,
    "scene_type": "action|dialogue|exposition|description",
    "required_knowledge": [
      "entity_type:entity_id.field"
    ],
    "pre_queries": [
      "SurrealQL query string"
    ],
    "post_writes": [
      "SurrealQL write operation"
    ],
    "consistency_checks": [
      "SurrealQL boolean expression"
    ],
    "voice_requirements": {
      "pov": "first|third_limited|third_omniscient",
      "tense": "past|present",
      "tone": "tone_description",
      "comparable": "comparable author/work"
    }
  }
}

This metadata structure enables the writer agent to:

• Gather necessary context via pre_queries
• Verify consistency before writing
• Write grounded in retrieved knowledge
• Store results with proper relationships via post_writes
• Flag gaps if required_knowledge is missing