leaf-work

Installation
SKILL.md

LEAF Work

Leaf before tree. Validate one cheap, inspectable instance before growing it into the whole artifact. The core move is never to generate the whole artifact upfront — that is how you produce confident-looking slop and lose the way before you can even tell the direction is wrong. It is to learn first, make one instance right, then expand — closing the cheapest decisive uncertainty before starting the next kind of work.

LEAF closes four kinds of uncertainty in order:

Phase What it makes you able to do Gates
Learn Judge what the work needs — learned, not guessed (run in leaf-learn; inherited here) ① Intent · ② Unknowns & Context
Example Prove one cheap instance right before scaling ③ Criteria · ④ Wireframe
Architect Generalize that instance into a shippable generator ⑤ Design · ⑥ Critic · ⑦ Tasks · ⑧ Artifact
Feedback Confirm it still holds, then settle what was established and what was learned ⑨ Review/sync · ⑩ Retrospect
Installs
4
Repository
hoetaek/leaf
GitHub Stars
3
First Seen
10 days ago
leaf-work — hoetaek/leaf