fable-skills (DizzyMii)
fable-skills
Six discipline skills for Claude Code: answer first, edits only as asked and a claim of success only with proof.
- 56 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The skills are instructions only. The installer copies folders into `~/.claude/skills` and with the flag edits the global `CLAUDE.md`.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/fable-skillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skills are instructions only. The installer copies folders into `~/.claude/skills` and with the flag edits the global `CLAUDE.md`.
This is third-party code. Review the repository files before installing.
What it does
Six skills pull the behaviour of Claude Code on Opus 4.8 toward the discipline of a stronger model. It concerns what the agent claims, when it stops, what it touches and how it reports. `fable-outcome-first` puts the answer first, `fable-finish-your-turn` tells it to finish reversible work without asking permission, and `fable-prove-it` allows claiming only the rung reached: written, runs or verified. `fable-scope-discipline` keeps the diff within the request and lists the rest separately, `fable-native-code` matches code to the file style and `fable-context-thrift` saves context during exploration. Each skill carries a rationalizations table: a real excuse from the model next to its rebuttal. The skills were pressure-tested on Opus 4.8 subagents, and the transcripts are in the repository.
Who it is for. Claude Code users on Opus 4.8 annoyed by sprawling diffs, needless questions and confident claims without checks.
Good fit when
- On a small edit the agent also rewrites neighbouring code and the diff balloons.
- After long exploration you want a direct yes-or-no answer rather than a retelling of the steps.
- The agent says done although tests never ran.
- You want new lines to match the file style, without defensive wrappers and narrating comments.
Not a fit when
- You work on another model: the skills are tuned to Opus 4.8 behaviour.
- You need better reasoning or long-horizon coherence: the skills change only discipline.
- The agent already behaves strictly and you do not want to touch `CLAUDE.md`.
Example request
Fix the bug in this file and report the result. Change only the bug and list the other issues without touching them.Limitations
The skills only address behaviour that instructions can shape: they do not improve reasoning or long-horizon coherence. They were tested on Opus 4.8 against a baseline of a `CLAUDE.md` with the superpowers pack; a bare model was not the baseline. The installer copies six folders into `~/.claude/skills/`, and the `--write-claude-md` flag additionally edits the global `CLAUDE.md`. Without the flag `CLAUDE.md` is untouched, and skills must then be called by name or the activation block added by hand.
How to disable. Delete the six `fable-*` folders from `~/.claude/skills/` and the marked activation block from `~/.claude/CLAUDE.md` if it was added.
FAQ
How do I install fable-skills?
Clone the repository and run `./install.sh`. On macOS and Linux add `--write-claude-md` to insert the activation block into `~/.claude/CLAUDE.md`. On Windows use `install.ps1` with the `-WriteClaudeMd` flag.
What is the activation block?
A short `claude-md-block.md` fragment. It ties each skill to a moment of work: task start, writing code, claiming a result and ending a turn.
What are the rungs in fable-prove-it?
The agent may claim only what it reached: written, runs or verified. State-changing commands go through the same check.
Does the set conflict with superpowers?
No. It serves as a calibration and reporting layer on top of superpowers and contradicts nothing in it.
Can I install it by hand?
Yes. Copy the six `skills/fable-*` folders into `~/.claude/skills/` and paste the contents of `claude-md-block.md` into `~/.claude/CLAUDE.md`.
Discussion
No questions or reviews yet.
Related
SkillsAn execution-discipline skill for Claude: staged planning, delegation to agents, failable checks and a double-check before delivery
A Claude Code plugin with four working-method skills: classify the task, act in steps, verify the result and generate domain adapters
A working folder for Claude Code where the agent asks questions and writes a new skill by Anthropic's guidance, not from memory.
11 skills for Claude Code: SKILL.md authoring, prompt tuning, refactoring, change review and planning in files.
A judgment layer for agents: rules for verifying results, checking before saying done and answering briefly, from real failures.
A Claude Code plugin that injects Anthropic guide rules into sessions, with the fable command expanding a vague ask
README in short
The README explains the idea of transferring discipline between models, lists the six skills by trigger moment and covers installation and a quick start. Test results and a comparison with superpowers follow.