autoresearch-coach-evolution

What problem does it solve?

This Skill automates the tedious, error-prone process of improving short coaching and microcopy strings so they become more concrete, emotionally engaging, and actionable without inventing facts or bypassing verification.

Core Features & Use Cases

• Mechanical scoring of individual coaching texts using grep/wc metrics (concreteness, jargon, brevity, actionability, emotion, voice) to create an objective baseline.
• Iterative generation of up to three constrained variants per target, selection by strict +5 mechanical improvement threshold, and automated verification steps including flutter gen-l10n and periodic flutter test runs.
• Use cases: optimizing in-app tips in localization ARB files, refining fallback or template messages in Dart services, and running batch sessions (20/30/50 attempts) to raise overall copy quality.

Quick Start

Run /autoresearch-coach-evolution 20 to run a 20-attempt session that scores, generates up to three variants, verifies mechanical improvement, and commits accepted edits.