BotShelf Vampire BOTSHELF VAMPIRE Register

RUN

Run it now

Everything on the shelf is text or code that runs in your own AI tool. BSV does not run it for you and does not set it up. Three moves are the same everywhere:

  1. Copy the full text. Use the Copy button on the item page, or download the file.
  2. Fill the blanks. Replace each bracketed placeholder, such as [GOAL], [INPUTS] or [CONSTRAINTS], with your real material. Leave no brackets behind.
  3. Keep the stop line. Most items say where the AI must stop: before sending, publishing, spending, deleting, pushing or placing an order. Leave that line in. The approval is yours to give.

Any chat AI

  1. Open a new chat.
  2. Paste the full text as the first message, with the blanks filled.
  3. Then send your material (the notes, the diff, the numbers) as the next message.
  4. Read the first answer against the item's own “stop” line. If it ran past the stop, say so and start a new chat.

ChatGPT

  1. One-off: new chat, paste the text as the first message.
  2. Reuse: create a Custom GPT and paste the text into Instructions, or put it in a Project's instructions.
  3. Fill the blanks before saving. Do not publish the GPT until you have tried it on your own material.

Grok

  1. Open Grok Bot and open a chat with the Bot that will do this job (or create a Bot for it).
  2. Fill the blank fields, send the text as a message, and ask the Bot to remember it. xAI's documentation says a personal Bot has no description and no Bot settings page: you change what it does by telling it in chat, and it keeps lasting rules in its memory.
  3. Team Bot: it does have a description, which its owner or a Manager edits under About in the Bot's details. We have not checked how much text it holds. If the text does not fit, send it in chat as in step 2.
  4. This follows xAI's documentation as we read it on 2026-11-18. We have not run it ourselves: on 2026-09-07 our own check reached an upgrade screen instead of Grok Bot on one account. Until 2026-11-18 pages here said to paste into a Bot's Description; that instruction was replaced. If your account cannot open Grok Bot, paste the same text as the first message of a normal Grok chat.

Claude Code

  1. One-off: start Claude Code in your repository and paste the text as the prompt.
  2. Standing rule: put the text in CLAUDE.md at the project root (or the project's instruction file).
  3. Let it read first. Ask for the plan, then the change. Review the diff yourself before any commit, push or merge.

Cursor

  1. One-off: paste the text into the agent chat with the files it should look at.
  2. Standing rule: add it as a project rule (a file under .cursor/rules/).
  3. Accept edits file by file. Run your tests before you keep them.

Gemini

  1. One-off: new chat, paste the text first, then your material.
  2. Reuse: make a Gem and paste the text into its instructions. In Google AI Studio, use the System instructions field.

DeepSeek, OpenRouter, Kimi and other API models

  1. Send the text as the system message. Send your material as the user message.
  2. Keep the API key in an environment variable. Never paste a key into the text.
  3. OpenRouter items name a routing idea, not a fixed model. Pick models you have access to and compare on the same input.

Kimi Code and other terminal coding agents

  1. Start the agent at the repository root and paste the text as the task.
  2. Work on a branch. Read every diff before you accept it.

Ollama, LM Studio, Open WebUI (local models)

  1. Ollama: when the item is a Modelfile, save it as Modelfile, run ollama create my-team -f Modelfile, then ollama run my-team.
  2. LM Studio: paste the text into the chat's System Prompt field.
  3. Open WebUI: paste it as the system prompt of a chat or of a model preset.
  4. Small local models skip rules more often. Check the first answer against the item's rules before you rely on it.

n8n, Dify, Flowise

  1. These Library items are instructions, not importable workflow files. Put the text into the system prompt of your LLM node (n8n, Flowise) or the app's instructions (Dify).
  2. Wire your own trigger and inputs. Put a manual approval step before anything that sends, posts or pays.
  3. Test with one harmless input before you connect real accounts.

CrewAI, LangGraph, AutoGen, multi-agent setups

  1. When the item is code, save it as a .py file, install the framework in a fresh virtual environment, and pin the version yourself.
  2. Point it at your model endpoint with environment variables. No keys in the file.
  3. When the item is a role layout (orchestrator, specialist, critic), use each role's text as that agent's system message.

MCP

  1. MCP items here are operating instructions for a setup you already have. They are not MCP servers.
  2. Use them in a client where your MCP servers are configured. Do not assume a server, credential or permission exists until you have checked it.
  3. Separate from those items: BSV itself can be read through a read-only MCP server. How to add it to your client.

Skills

  1. Skill texts on the shelf are plain instructions. Paste one into any chat when the job comes up.
  2. To keep one as a Claude Code skill, create .claude/skills/<name>/SKILL.md, write the YAML front matter (name and description) yourself, and paste the text under it. The shelf text does not include that front matter.
  3. The Grok Bot tools are different: their skills are complete SKILL.md files with the front matter included, for Grok Build (.grok/skills/) and for pasting into Grok Bot.

TradingView, MT4, MT5 and other chart platforms

  1. Open the tool's page and follow its “Install, inspect, then test” steps. Keep LICENSE.txt with the source.
  2. Guides: Pine · MetaTrader · Testing · Licenses.
  3. Replay and demo first. This is analysis and tooling, not investment advice, and nothing here places orders for you.

When it does not work

Find something to run Fork guide What is open