- Home
- Skills
- Prompt & Skill Engineering
- LLMwiki4roleplay
Works with the AI Agents you already use
LLMwiki4roleplay
Automate novel-to-Obsidian knowledge graph construction and drive lore-accurate character roleplay.
Free
LLMwiki4roleplay
Example session with this skill installed
Using the LLMwiki4roleplay skill, split and ingest fantastic_beasts.txt into ./roleplay/vaults/fb/, and start processing the first batch of chapters.
- Read your context and instructions
- Compiled the llmwiki4roleplay
I have successfully run the split-book command on fantastic_beasts.txt and initialized the vault at ./roleplay/vaults/fb/. I am now executing the ingestion loop for the first batch (Chapters 1-3).
📊 Ingestion Progress
- Source:
fantastic_beasts.txt - Output Vault:
./roleplay/vaults/fb/ - Progress saved:
.progressfile updated (Ready to resume anytime). - Indexing: Global index
index.mdupdated. Team indices initialized.
🕸️ New Entities Extracted (First Batch)
I have identified 2 new characters and 1 creature, generating their Markdown files with embedded [[wikilinks]] to form the knowledge graph:
| Entity ID | Type | Aliases |
|---|---|---|
newt-scamander | Character | Newt, Mr. Scamander |
jacob-kowalski | Character | Jacob, Mr. Kowalski |
niffler | Creature | The Niffler |
📄 File Preview: newt-scamander.md
---
id: newt-scamander
name: Newt Scamander
aliases: [Newt, Mr. Scamander]
type: character
---
# Newt Scamander
Newt is a British Magizoologist who arrives in New York with a magical suitcase. He encounters [[jacob-kowalski]], a No-Maj (Muggle), whose accidental release of a [[niffler]] causes chaos in the city. *(Note: Team index `_index-newt-scamander.md` initialized for his extended lore.)*
🛡️ Roleplay Engine Test (Dual-Memory & Boundary)
Let's simulate a chat session to demonstrate the system.
Command: roleplay-context --character newt-scamander
System Initialization:
Long-term Memory: Loads newt-scamander.md + _index-newt-scamander.md (Canonical facts).
Short-term Memory: Loads history.txt from roleplay/memories/fb/newt-scamander/.
Knowledge Boundary Enforced: entity-context --reader newt-scamander filters out any lore he hasn't encountered yet.
Simulated Interaction:
User: "Newt, do you know what Grindelwald is planning right now?"
Newt: "I'm afraid I haven't the faintest idea. I'm just trying to track down my escaped [[niffler]] before MACUSA finds out. I really don't want to get involved in any of that."
(Notice how he references his own known entity but strictly adheres to his knowledge boundary, refusing to reveal plot points he hasn't encountered in the current timeline.)
⏭️ Next Steps
Continue ingesting Chapters 4-6.
Process new locations and link them to existing entity profiles (e.g., macusa-hq).
Memory Management: The dual-memory system is active. As you chat with Newt, short-term conversations will be stored in history.txt. Once it exceeds the threshold (200 topics), the system will automatically use cleanup-memories (LRU cache logic) to consolidate memories into the Memory Vault while protecting key states.
Type "Continue" to resume from the .progress breakpoint.
Connects securely to your tools. The creator never sees your data.
About this skill
Turn raw novels into an interactive, lore-accurate Obsidian knowledge graph & roleplay engine.
Building a comprehensive encyclopedia or roleplaying engine from raw text files is a labor-intensive manual process. This skill automates the extraction of entities and relationships from novels, constructing an incremental knowledge graph in Obsidian that powers lore-accurate character interactions.
✨ What it does
- Incremental Graph Construction: Processes text chapter-by-chapter to build an Obsidian vault of people, locations, items, and events.
- Strict Knowledge Boundaries: Prevents OOC (Out of Character) responses. Characters only know what they have actually encountered in their own narrative, rather than the AI reading the whole book and revealing spoilers.
- Dual-Memory Roleplay: Separates canonical book facts (long-term memory) from your conversation history (short-term memory). Characters remember your past interactions without breaking their lore.
- Scalable Indexing: Uses LRU cache logic and hierarchical indices (
index.md/_index-{parent}.md) to handle massive texts without slowing down retrieval or duplicating entities.
⚙️ How it works (CLI Toolset)
split-book: Partition a raw.txtfile into chapters and generate an ingestion plan.ingest: Run the ingestion loop to extract entities, create Markdown files, and log progress. Supports resuming from breakpoints (.progress).roleplay-context: Retrieve a character's "long-term memory" from the book vault and "short-term memory" from chat logs to initialize a session.search-entities: Query the index to find specific lore or verify aliases before creating new records.
🚀 Why this beats prompting it yourself
Generic prompts lose track of thousands of entities and specific page references over time. This skill enforces a strict knowledge boundary and file-based persistence, ensuring the AI only knows what is actually in the text or the chat history.
💡 Use Cases
- Build a searchable wiki for a complex fantasy novel series.
- Create an interactive roleplay bot that remembers previous player interactions.
- Audit character arcs and location usage across long-form manuscripts.
- Summarize long-running web novels into a structured knowledge base.
⚠️ Requirements & Known Limitations
- Environment: Requires Python 3.10+,
pyyaml, and Claude Code (or compatible agents). - File System: Requires local file IO to generate the Obsidian vault structure.
- Command Line: You must have basic familiarity with terminal commands to run the Python CLI tools.
- Language: Supports Chinese by default for prompts and narrative extraction (roleplay adapts to the user's language).
Note: This is the open-source version. For full documentation, source code, and setup instructions, please visit the GitHub Repository (English README).
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Prompt & Skill Engineering.
Defaults to Chinese for internal prompts. Requires specific directory layout for tools. Maximum of 200 memory entities before LRU cleanup triggers.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 27 days ago
- Free with an Agensi account
- Software
Python 3.10+Claude CodePyYAML