kgmkm/novel2hermes_jp — explained in plain English
Analysis updated 2026-05-18
Plan a long Japanese novel through a structured proposal, character, and plot workflow.
Catch plot contradictions and continuity errors using searchable story memory.
Revise a manuscript by comparing checks from several different LLMs at once.
Generate consistent illustration prompts for a novel's characters and scenes.
| kgmkm/novel2hermes_jp | aerdelan/housand-domaintoolmatrix | afomera/apple-foundation-models-ruby-sdk | |
|---|---|---|---|
| Stars | 22 | 22 | 22 |
| Language | — | TypeScript | Ruby |
| Setup difficulty | moderate | easy | moderate |
| Complexity | 3/5 | 2/5 | 3/5 |
| Audience | writer | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires Hermes Agent v2.x and the vecmemori memory provider already set up before installing this skill.
This repository is a skill package for Hermes Agent, an AI agent framework, designed to help write long Japanese novels using an external memory tool called vecmemori. The idea is to split fiction writing into planning, drafting, and revision stages, similar to how an editor and a novelist would divide the work. The planning stage builds a proposal, detailed character sheets, a setting description, and a plot outline. The character sheets use a 36 item template meant to also feed image generation tools such as ComfyUI, so a character's appearance can be described consistently enough to illustrate later. Once planning is done, the writer moves into drafting chapters, then into a revision stage that uses several different large language models, each one checking the manuscript from a different angle: logic, writing style, historical or setting accuracy, and how the story reads from a reader's point of view. Using different models for each check is meant to avoid the bias that comes from relying on just one model's judgment. vecmemori is the piece that gives the system long term memory. It stores characters, world details, and foreshadowing, and supports Japanese language semantic search so past chapters can be searched and checked for contradictions as the story grows. The skill also includes support for generating illustrations, working with tools like GPT Image, Nano Banana, or ComfyUI, from picking which scene to illustrate through suggesting a prompt, with an optional human review step when using ComfyUI. Setup requires Hermes Agent version 2.x already installed along with the vecmemori memory provider and at least one LLM provider such as OpenRouter, Nous Portal, Anthropic, or OpenAI. Installing the skill itself is a matter of telling Hermes Agent the repository URL, or manually cloning it and linking it into the Hermes skills folder. The project works with any fiction genre, including fantasy, science fiction, mystery, romance, and horror, and is released under the MIT license.
A skill set for the Hermes AI agent that plans, writes, and revises long Japanese novels using persistent memory search and multi model revision.
MIT license: use freely for any purpose, including commercial use, as long as you keep the copyright notice.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly writer.
This repo across BitVibe Labs
Verify against the repo before relying on details.