Last commit on July 15, 2026·Created on July 6, 2026

Sahir619/fable-method

“A structured framework for implementing think-act-prove workflows and adversarial verification.”

Combined rank

#168

across all skills

In Agent Orchestration

#9

category rank

Stars

2.3k

+0.3% in last 7d

Forks

326

-0.3% in last 7d

Watchers

8

0.0% in last 7d

Traction score

TL;DR

This repository provides a method for building trusted skill bundles through research-backed workflows and flowcharts. It helps AI agent builders move from simple generation to a structured process of domain research and evidence-based verification.

The system emphasizes adversarial verification, treating completed work as a set of claims that must be re-verified to deliver an evidence-based verdict on the output.

WHO IT'S FOR

AI agent / automation builders

implementing structured think-act-prove workflows

QA / test automation engineers

performing adversarial verification of AI outputs

Technical leads / architects

standardizing AI coding rules and evaluations

Backend / infrastructure engineers

managing immutable configuration release cycles

Categories

Agent OrchestrationPrimary

Software Engineering Testing

Compatible Agents

Claude Codex Cursor

Claude Code plugin

Installs: Complete versioned `fable` plugin containing all four Fable Workflow skills: fable-method, fable-loop, fable-judge, and fable-domain. · Claude Code

Before you start

Claude Code with plugin support

Inside a Claude Code session, add the repository as a plugin marketplace.
```
/plugin marketplace add Sahir619/fable-method
```
In the same Claude Code session, install the `fable` plugin from the `fable-method` marketplace.
```
/plugin install fable@fable-method
```

.claude-plugin/plugin.json ↗ · Checked Sep 18, 2026.claude-plugin/marketplace.json ↗ · Checked Sep 18, 2026README.md ↗ · Checked Sep 18, 2026

Give it something to do.

Suggested first task

Plan a task

Uses fable-method

Replace `<task>` with the task, relevant repository path, constraints, and desired outcome.

Use the fable-method skill in plan mode for <task>. Gather evidence, recommend one approach, and stop without changing files.

skills/fable-method/SKILL.md ↗ · Checked Sep 18, 2026

Suggested first task

Verify completed work

Uses fable-judge

Replace `<completed work>` with the recent work, diff, directory, branch, or pasted report to judge.
Provide a runnable local environment for claims involving tests, builds, scripts, or pages.

Use the fable-judge skill to verify <completed work> without making changes. Re-run its claimed checks and give an evidence-based verdict.

skills/fable-judge/SKILL.md ↗ · Checked Sep 18, 2026

Suggested first task

Fix a local failure

Uses fable-loop

Replace `<local failure>` with a failing test or reproducible bug.
Replace `<repository path>` with the local project directory.
Provide the failing command or reproduction steps if the project does not contain them.

Use the fable-loop skill to investigate and fix <local failure> in <repository path>. Make the smallest correct change and verify the result.

skills/fable-loop/SKILL.md ↗ · Checked Sep 18, 2026