cozytab
Official@cozytab
cozytab provides fable-mode, a work-discipline protocol that elevates non-frontier model output to Fable-5-grade quality via six orchestration levers.
Agent Skills by cozytab
Showing 1 vetted skills indexed across 1 GitHub repositories.
Frequently Asked Questions About cozytab
FAQPage SchemaWhat tasks can I accomplish with cozytab's fable-mode skill?▼
Fable-mode raises output quality of Opus 4.8 or any non-frontier model to Fable-5-grade by applying six levers: plan gate, small-card execution, adversarial self-check, real-product verification, context hygiene, and checkpoint autonomy. It trades extra orchestration for single-pass quality.
Who should use the fable-mode skill?▼
Engineers and practitioners running non-frontier models who need frontier-grade, single-pass results on quality-sensitive work. It suits users willing to spend extra orchestration effort in exchange for disciplined planning, self-verification, and clean context management.
How is fable-mode activated during a session?▼
It activates only on explicit request, such as 'use fable mode', 'work like Fable 5', 'rigorous mode', or the Chinese '严谨模式', or automatically when the project contains a .fable/ directory whose hooks handle arming. It never auto-triggers on task size or generic quality phrases.
Does fable-mode turn on automatically for important tasks?▼
No. It deliberately does not auto-trigger merely because a task is large, important, or quality-sensitive, nor on vague phrases like 'do it well'. If a task looks like it would benefit, the assistant asks the user whether to enable fable-mode instead of assuming.
What are the prerequisites for using fable-mode?▼
You need a model session such as Opus 4.8 or another non-frontier model, plus either an explicit activation phrase or a project pre-armed with a .fable/ directory. No other dependencies are specified in the manifest.