ultraprompt

This plugin provides twelve strategy skills (eight core axes plus four draft candidates) distilled from reasoning traces of Claude Fable 5.1, designed to improve how Opus, GPT, and open-weight coding agents explore codebases, manage hypotheses, verify solutions, articulate trade-offs, enumerate failures, self-correct, maintain spec fidelity, and ensure incremental safety. The skills are portable English prompts grounded in evidence from real Claude Code sessions across diverse domains like frontend design, backend systems, ML training, kernel optimization, and Linux syscalls.