Skip to content

RamiAltai/fableify

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

✨ fableify ✨

Your AI is smart. This makes it careful.

fableify — Fable's working habits as plain-text cards; deal them to any model to fableify it

Give any AI assistant the working habits of a top-tier engineer, as plain text files you can read, in about the time it takes to copy-paste.


The problem

You've noticed it. You ask an AI to do something and it says "Done! ✅" — but it never actually checked. It writes fifty lines where five would do. It fixes the thing you pointed at and quietly leaves three identical bugs next to it. It agrees with you a little too eagerly.

The expensive, top-tier models do this less. But you don't always want to pay top-tier prices for every little task.

The idea

fableify is a stack of "habit cards" for your AI.

fableify — Fable's working habits as plain-text cards; deal them to any model to fableify it

Each card is a short, plain-text file that teaches one good habit — the kind a seasoned professional has and a hurried one doesn't. You hand a card to your AI, and it starts behaving like the careful, senior version of itself:

  • 📖 Reads before it writes — understands the whole problem instead of pattern-matching the first line.
  • Verifies before it says "done" — actually runs the check, instead of claiming it passed.
  • 🚫 Refuses to fake a green checkmark — tells you when something failed, plainly.
  • ✂️ Does less — the smallest solution that actually works, not the most impressive-looking one.
  • 🎯 Says the useful thing first — the answer, then the detail. No throat-clearing.

The trick: these habits don't need the smartest model. Written down clearly, a cheaper model can follow them — so you get careful, senior-grade work at that model's price.

Why you'd use it

You want the quality of the best AI without the bill. Load a habit card and a mid-tier model works like a careful one.

You're tired of "Done! ✅" that wasn't. The whole point of these cards is an AI that checks its work and tells you the truth about it.

You want to see and tweak the rules. No black box. Every habit is a short file in plain English. Don't like a rule? Edit it. It's yours.

No install, no account, no lock-in. It's just text. Copy the card into whatever AI tool you already use.

What's in the box

Seventeen habits, grouped by what they're for. A few highlights:

If you want your AI to… Reach for
Stop claiming things work without checking verification
Fix the real bug, not just the symptom debugging, coding
Catch security holes and prove they're real security-review
Actually push back instead of flattering you sparring-partner
Think in real options with honest confidence pure-thinking
Write clearly and cut the fluff writing
Plan risky work before diving in planning
Design something clean instead of over-built design-system, design-ui
Give straight career advice, not a pep talk career-advisor

…plus habits for reviewing code, handing off work, making estimates, and orchestrating bigger jobs. The full list with a one-line description each is in docs.md.

How to use it (about 30 seconds)

  1. Pick a card — open the skills/ folder and find the habit you want (say, skills/verification/).
  2. Hand it to your AI — paste the card's SKILL.md into your AI's system prompt / custom-instructions box. That's it. The habit is now on.

Using Claude Code? Drop the skills/ folder into your project and it picks them up automatically. Full setup notes for every tool are in docs.md.

Does it actually work?

Yes — and we can show our work. Every habit card ships with three built-in tests: one that proves it kicks in when it should, one that proves it stays quiet when it shouldn't (so it doesn't turn every question into a lecture), and one tricky edge case.

The whole catalog was also stress-tested by an independent reviewer whose only job was to break the claim that it's ready — checking every card against the rules and running the behavioral tests. It passed with zero hard failures.

The name

To fableify something is to give it the good habits of Claude Fable, Anthropic's top model — and hand them to any other model as a fable it can follow. The strong model's judgment, written down so a cheaper one can borrow it.

Bring your own AI

These are plain markdown files, not code. They work with any assistant that lets you set instructions — Claude, and most others. No dependencies, nothing to run, nothing to sign up for.


Careful beats clever. Now your AI can be both.

📚 Full documentation & the complete skill list →

About

No description, website, or topics provided.

Resources

License

Stars

0 stars

Watchers

0 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors