ruvnet/Bot-Generator-Bot

The ultimate bot generator bot prompt. Use this prompt to create powerful ChatGPT bots for anything you can imagine.

What it solves

This tool solves the problem of turning a rough bot idea into a working agent without ad-hoc prompt tinkering. Instead of relying on a loose collection of prompts, it provides a compiler that turns a structured specification into a fixed, verifiable manifest, and a runtime that runs that agent safely under operator-controlled limits. It also keeps model access and tools strictly gated, so nothing runs outside the boundaries you set.

How it works

You define a bot by writing a JSON spec with fields like name, purpose, context, examples, output schema, and allowed tools. The compiler checks that spec against strict rules (field limits, type checks, extra-field rejection) and produces a deterministic manifest with a consistent SHA256. That manifest can then be run through a CLI or an MCP server. The runtime uses the MetaHarness agent kernel, private RuVector memory for retrieval, and optionally RuFlo for keyword routing. Any model call requires you, the operator, to explicitly enable it and supply credentials via environment variables — nothing is embedded in the spec. The compiler itself works fully offline, and the runtime has bounded turns, tool calls, tokens, and output size. You can also export the manifest as a standalone project with its own CLI and MCP setup.

Who it's for

This is for developers who want to generate structured, typed agents from a spec and run them in a controlled way. It's especially useful if you use Claude Code, Codex, or an MCP-compatible host, since it ships installable skills and a local MCP server. It's also for people who need reproducible agent builds — the deterministic compiler means the same spec always produces the same manifest, which helps with auditing and tracking.

Highlights

  • All 39 original 2023 prompts migrated into typed agent templates, skills, and MCP builders
  • Deterministic compiler that produces identical manifests and SHA256 hashes from the same spec
  • Strict typed output validation — rejects missing fields, extra fields, wrong types, and unsafe numbers
  • Operator tool policy: requests must match a fixed catalog and an operator allowlist; requests never grant authority
  • CLI and MCP server offering 50 tools, 42 prompt templates, and three resources
  • Runnable agents with a bounded model-and-tool loop, typed final output, and standalone project export
  • Private manifest-scoped RuVector memory with native vector search
  • Optional RuFlo keyword routing as a pinned upstream hook
  • Uses the actual MetaHarness kernel with host adapters, session replay, and authenticated field memory
  • Autogenous pinned upstream fitness gates with no automatic promotion
  • CI delivery with tests, dependency audit, provenance checks, and downloadable artifacts

Related

  • Project
  • Project
  • Project
  • Project