writer-memory_skill

Overview

This skill is an agentic memory system for fiction writers that persistently tracks characters, relationships, scenes, worldbuilding, and themes. It keeps a git-friendly JSON memory file so writing teams or solo authors can maintain consistent voice, emotional arcs, and scene purpose across sessions. It includes validation, natural-language queries, and emotion-focused synopsis generation.

How this skill works

The system stores structured memory in a single .writer-memory/memory.json and auto-backups before changes. Writers add or update character, relationship, scene, world, and theme entries via simple commands; the skill synthesizes context for queries, validates dialogue against a character's voice, and generates an emotional synopsis from aggregated memory. It supports session loading so agents use the same context across runs.

When to use it

• Keeping character voice and emotional arcs consistent across drafts
• Tracking relationship evolution and key events after each scene
• Validating dialogue to match speech level, tone, keywords, and taboos
• Generating an emotion-focused synopsis to check thematic coherence
• Querying the current emotional state or likely reactions during writing

Best practices

• Start by building character memories (arc, tone, speech level, keywords, taboos)
• Update relationships and scenes immediately after key plot beats
• Run dialogue validation while drafting to catch voice drift early
• Generate periodic synopses to ensure themes remain coherent
• Keep memory.json under version control and create backups before big edits

Example use cases

• Add a new character and define arc, tone, speechLevel, and taboo words before drafting their scenes
• Record a rooftop reunion scene and update the relationship timeline with the scene catalyst
• Validate a tough emotional line to confirm it fits the character’s established tone
• Ask a natural-language query like “What would the protagonist say in a breakup?” to get context-aware suggestions
• Export the full memory to markdown for collaborator review or editorial notes