Outliner Agent
The Outliner Agent creates comprehensive chapter-by-chapter outlines for your book project, providing a detailed roadmap for the writing process.
Overview
The Outliner Agent takes the initial book concept and expands it into a structured outline, breaking down the story into chapters and key scenes. It ensures proper pacing, plot development, and narrative structure.
Features
Chapter Structure
- Creates detailed chapter summaries
- Establishes clear story beats
- Maintains proper pacing
- Ensures narrative flow
Plot Development
- Develops main plot and subplots
- Places major plot points
- Tracks character arcs
- Maintains story tension
Content Organization
- Generates chapter titles
- Creates scene breakdowns
- Includes key story elements
- Tags important plot points
Output Format
The outline is generated in Markdown format:
# Chapter 1: Title
- Opening scene description
- Key plot points
- Character interactions
- Scene objectives
# Chapter 2: Title
...
Integration
The Outliner works closely with:
- Concept Generator for initial story direction
- Character Generator for character arc integration
- Worldbuilding Agent for setting integration
- Chapter Writer for detailed chapter development
Usage
# Through the CLI
libriscribe outline
The Outliner is also run automatically during guided setup after concept generation.
Best Practices
-
Review and Iteration
- Review the generated outline
- Make necessary adjustments
- Ensure all plot threads are addressed
- Verify pacing and structure
-
Chapter Balance
- Check chapter lengths
- Verify plot point distribution
- Ensure proper character focus
- Maintain narrative momentum
-
Integration Points
- Mark character development moments
- Note worldbuilding elements
- Identify research needs
- Tag emotional beats
Error Handling
The Outliner Agent includes error handling for:
- Invalid project data
- File system operations
- Content generation issues
- Format validation
Pipeline Context Propagation
These behaviours were added to ensure every user answer reaches the outline.
User Chapter Count Respected
_get_max_chapters() now reads num_chapters_str from the PKB to detect whether the user explicitly set a chapter count. When set, the user's preference is enforced (capped by length tier: Short Story ≤ 2, Novella ≤ 8, Novel ≤ 20). When not set (default num_chapters=1), the length-tier ceiling alone governs — the default is never mistaken for a user preference.
A chapter count hint is also injected into the LLM prompt so the model targets the right number.
Author Notes Injected Into Prompts
Both execute() (outline-level) and generate_scene_outline() (scene-level) append an Author notes block when any of these Advanced-mode PKB fields are set:
| Field | Label |
|---|---|
inspired_by | Inspired by |
author_experience | Author experience |
key_takeaways | Key takeaways |
case_studies | Includes case studies |
actionable_advice | Includes actionable advice |
marketing_focus | Marketing focus |
sales_focus | Sales focus |
research_question | Research question |
hypothesis | Hypothesis |
methodology | Methodology |
Dynamic Questions Propagated
Genre-specific Q&A collected during Advanced setup (pkb.dynamic_questions) is appended to the Book Description in generate_scene_outline() so scene-level prompts reflect the author's genre-specific intentions.
Chapter Limit Enforced
_enforce_chapter_limit() is called after process_outline() to strip any extra chapters the LLM may have generated beyond max_chapters.
Description Never Overwritten
The user's original book description is never replaced by any LLM-generated summary. The internal book_summary_lines extraction loop still runs for parsing purposes but the assignment project_knowledge_base.description = … was removed.