Oh My Fable — AI skill for Claude Code
oh-my-fable — Fable 5's way of working a long task (plan first, self-correct, never lose the thread), as a model-agnostic agent harness.
How to install Oh My Fable
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open didrod205/oh-my-fable and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Oh My Fable does
oh-my-fable — Fable 5's way of working a long task (plan first, self-correct, never lose the thread), as a model-agnostic agent harness. The run lives in one serializable RunContext, checkpointed every step, so a crash is a pause. Zero deps, deterministically testable.
Alternatives in AI
- CLI WeChat Bridge — 将 AI 命令行工具以最为原生的方式,集成到微信 ClawBot / 企业微信 bot 中,目前支持集成 Codex、Claude Code、OpenCode、Pi Agent 478 ★
- Claudexor — Multi-harness control plane for Claude Code, Codex, Cursor, and OpenCode: quota-aware rotation across multiple 425 ★
- Ralph Loop — A long-running AI agent loop 292 ★
README
oh-my-fable
Fable 5's way of working a long task — plan first, self-correct every step, never lose the thread — as a model-agnostic agent harness.
The fable is Fable 5's way of thinking; the oh-my- is because, like oh-my-zsh, you just want the good defaults. The mindset is the model's — the engine is any provider.
[](https://www.npmjs.com/package/oh-my-fable) [](https://github.com/didrod205/oh-my-fable/actions/workflows/ci.yml) [](https://www.npmjs.com/package/oh-my-fable) [](https://www.npmjs.com/package/oh-my-fable?activeTab=dependencies) [](./LICENSE)
npx oh-my-fable demo
**Watch a crash → resume in 10 seconds — no API key, no setup.**
Already paying for Claude Code? Then this costs $0 per token
npx oh-my-fable run "refactor utils.ts and run the tests" --provider claude --cli-tools
Drives Claude Code as a **durable, tool-using agent** on the login you already have — **no separate API key, no per-token billing.** Claude edits the files and runs the commands itself; oh-my-fable stays the durable planner/reflector around it, checkpointing every step so `resume` always works. ([all the CLI options ↓](#or-use-it-from-the-terminal))
Symptom → what stops it
| you have seen this | what stops it |
|---|---|
| The process dies at step 47 of 60 — and the agent starts over from zero. | `RunContext` is checkpointed after every step; `resume()` continues at 47 → |
| It loops on the same step forever, re-trying the same broken approach. | [maxStepAttempts marks th |
Related Skills
Fuck It We Ball
🏀 Claude Code skill: runs an approved plan end to end with subagents — ordered by urgency × unblocking × seve
Slack AI Bot
Agent Skill: build a two-way Slack AI bot — mentions, DMs and thread follow-ups answered by a model with user-
Memctx
Autonomous session memory for Claude Code. Never lose context, never repeat yourself. AI-powered summaries, se
Terminal Workspace Snapshot
Never lose your Claude Code or Codex tabs again. Snapshot and restore AI coding sessions as color-coded Window
Unsent
AutoRecover for your AI agent prompts: never lose a half-written Claude Code message to a crash, a closed wind
Skill Wrangler
Skill Wrangler is a cross-platform desktop app for anyone working with Agent Skills — the SKILL.md folders use
Related Agents
Opencode Generalist
OpenCode, the open-source model-agnostic CLI harness, for general coding
Vzt Planner
Frontier planning agent (Fable 5.1) — the LAST rung, for planning with no prior art to reason from: novel or g
Seam Auditor
Composition adversary - finds two correct units whose join lies. Orthogonal to the house happy/sad-path signat