DannyMac180/fable-advisor

Claude Fable as an orchestrator for Opus, GPT and Grok

What it solves

Fable Advisor implements an "architect pattern" for AI-driven software development, preventing the common issue of a single model's blind spots and high token costs. It separates the high-level design and final verification from the actual code implementation, ensuring that code is reviewed by a different model family than the one that wrote it.

How it works

The project is a plugin for Claude Code that organizes work into three distinct "lanes" based on the task complexity and role:

  • Architect (The Session): Acts as the project lead. It handles requirements, decomposition, and specifications, and routes tasks to the appropriate implementation lane.
  • Implementation Lanes (Codex): These are cross-vendor lanes that handle the actual typing of code.
    • Routine Lane: Uses GPT-6 Luna for standard tasks where the spec is fully determined.
    • High-complexity Lane: Uses GPT-6 Sol for tasks requiring judgment, such as hard debugging or security-sensitive paths.
  • Review Lane (Fable 5.1): A read-only skeptic that reviews the final accumulated changes in a clean context before any deliverable is marked as done.

Who it’s for

Developers using Claude Code who want a multi-model orchestration layer to improve code quality through cross-vendor review and optimize token spending by using cheaper models for routine implementation.

Highlights

  • Cross-Vendor Review: Built-in diversity by using Claude models for architecture and review, and OpenAI models for implementation.
  • Orchestration Skill: Provides a routing doctrine and a six-part spec contract to make context-free delegation safe.
  • Customizable Lanes: Users can configure which models and reasoning effort levels (low to ultra) are used for each lane via /fable-advisor:setup.
  • Advisor-only Mode: A lightweight option to use the Fable 5.1 reviewer without the full orchestration layer.

Related

  • Project
    duolahypercho/fusion-fable

    Fusion‑Fable is a Claude Code skill that runs a prompt in parallel on multiple LLMs (Opus 4.8, GPT‑5.5, Gemini 3.1 Pro), lets Opus 4.8 judge the independent answers, and synthesises a final, structured response. It installs as a Claude skill with slash commands, saves provenance files, and includes an iterative planning mode (`/fusion‑plan`) that integrates with the OMC planning system. Zero‑setup panel (two Opus 4.8 runs) works out‑of‑the‑box; richer panels need the `codex` and `agy` CLIs. The trade‑off is higher token cost and latency, but the result is a higher‑quality, auditable answer.

  • Dispatch
    Claude 5 Fable: Evaluating the Mythos-Class AI Model

    Claude 5 Fable represents a significant leap in AI capability, shifting the human role from active steering to high-level commissioning of complex, multi-hour autonomous workflows.

  • Dispatch
    Fable 5 vs. GPT-5.6 Sol: Benchmarking NP-Hard Optimization with /goal Mode

    A benchmark of Claude Fable 5 and GPT-5.6 Sol on an NP-hard fiber-network design problem reveals that Fable 5 is significantly more consistent and effective, while the /goal feature provides inconsistent benefits that can actually degrade average performance.

  • Project
    codejunkie99/fable-orchestrator

    A local-first routing skill for Codex that separates high-level planning and adjudication by Claude Fable 5.1 from code implementation by specialized OpenCode Go agents.

  • Project
    Chachamaru127/claude-code-harness

    A management framework for AI coding assistants that adds structured planning, safety guardrails, and mandatory review steps to ensure reliable code generation.