</>MCP Agents Market
Plugin

fable-method

by Sahir6192.3kPythonUpdated 2026-07-15

The Fable Workflow: how Claude Fable 5 worked, distilled into skills any model can run, with the eval that keeps it honest. Think / act / prove.

Claude CodeCodexCursoraider

The Fable Method is a Claude Code plugin that implements a structured think-act-prove workflow for AI agents, distilled from Claude Fable 5's problem-solving approach. It provides four interconnected skills (fable-method, fable-loop, fable-judge, fable-domain) that guide models through evidence gathering, decision-making, verification, and honest reporting. The plugin includes domain-specific adapters for marketing, research, data analysis, business operations, and other sectors beyond coding. Extensively tested across 15 evaluation rounds with 260+ agent runs, it demonstrably improves mid-tier model performance on complex tasks requiring careful verification and authority conflict resolution.

Key Features

Seven-step workflow (classify, define done, gather evidence, decide, act, verify, report) with hard bounds and escape hatches to prevent runaway agent behavior
fable-judge: adversarial verification that re-runs claimed checks, diffs actual changes, and hunts for false completion claims and weakened tests
fable-loop: orchestrated execution with parallel evidence subagents, approval gates for ambiguous or irreversible actions, and adversarial verifiers
fable-domain: generates new domain adapter bundles (adapter, trap fixture, smoke eval) for specialized sectors following an observed research-driven process
Eight built-in domain adapters (marketing, research, data, business, finance, legal, design, devops) with binding minimum evidence requirements
Comprehensive evaluation suite with 14 trap fixtures testing spec conflicts, fraudulent completion reports, unauthorized actions, and cross-tier model performance
Installable as versioned Claude Code plugin via marketplace or as standalone skills for Codex, Cursor, aider, and other agents
Proactive integration via global CLAUDE.md configuration for automatic application on multi-step tasks

Use Cases

  • 01Improving reliability of unattended or multi-step AI agent tasks by enforcing structured evidence gathering and verification loops
  • 02Catching specification vs. test conflicts where an agent might silently fix correct code to match a wrong test
  • 03Verifying completed agent work for false completion claims, weakened tests, or skipped steps before presenting results
  • 04Boosting mid-tier model performance on complex tasks that frontier models handle natively (inversely proportional tier lift)
  • 05Extending structured workflows to non-coding domains like marketing copy review, research projects, data analysis, and business decisions
  • 06Generating trustworthy domain-specific adapters with researched evidence requirements and adversarial test fixtures

Related Plugins

View more

andrej-karpathy-skills

A single CLAUDE.md file to improve Claude Code behavior, derived from Andrej Karpathy's observations on LLM coding pitfalls.

Plugin216.6k

claude-mem

Persistent Context Across Sessions for Every Agent – Captures everything your agent does during sessions, compresses it with AI, and injects relevant context back into future sessions. Works with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, OpenCode + More

Plugin95.2k

Understand-Anything

Graphs that teach > graphs that impress. Turn any code into an interactive knowledge graph you can explore, search, and ask questions about. Works with Claude Code, Codex, Cursor, Copilot, Gemini CLI, and more.

Plugin85k

rtk

CLI proxy that reduces LLM token consumption by 60-90% on common dev commands. Single Rust binary, zero dependencies

Plugin82.2k

fable-method — FAQ

What is the Fable Method AI sub-agent?+

The Fable Method is a Claude Code plugin that provides four structured workflow skills distilled from Claude Fable 5's problem-solving approach. It guides AI agents through evidence-based thinking, acting, verification, and honest reporting with hard bounds to prevent common failure modes like false completion claims and authority conflicts.

How do I install the fable-method plugin in Claude Code?+

Inside any Claude Code session, run '/plugin marketplace add Sahir619/fable-method' then '/plugin install fable@fable-method'. All four skills arrive namespaced and versioned. For standalone installation in other agents, clone the repo and run install.sh (Unix) or install.ps1 (Windows).

Which AI clients and agents work with the Fable Method?+

The plugin is designed for Claude Code with marketplace installation. The standalone skills work with Codex, Cursor, aider, or any agent that accepts system prompts via the AGENTS.md file included in the repository.

Do I need API keys or prerequisites to use fable-method?+

No external API keys are required. You need a working Claude Code installation for plugin use, or an agent framework that supports custom skills for standalone installation.

Is the Fable Method free to use?+

Yes, the Fable Method is open source under the MIT license and free to use, modify, and distribute.

What's the difference between fable-method, fable-loop, fable-judge, and fable-domain?+

fable-method is the core workflow applied inline; fable-loop orchestrates full parallel execution with approval gates; fable-judge adversarially verifies completed work; and fable-domain generates new domain adapter bundles with research, traps, and evals.

How do I install fable-method?+

Open the source repository on GitHub and follow its README. fable-method is a plugin — MCP Agents Market links you directly to the official repo.

Is fable-method free?+

fable-method is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.

Related searches

Developer Tools PluginsProductivity & Workflow PluginsTesting PluginsBest MCP ServersBest SkillsBest AgentsTop rated listingsClaude Code toolsBrowse all categories