Score every skill edit by rolling the task back and re-running it.

Skill-α (arXiv 2608.01678) reframes skill generation as RL over sequential edits, decomposing skill construction into individually evaluable changes. The novel signal is a rollback reward that scores each modification by comparing downstream task execution using the original skill against the edited one, which solves the core problem that skills have no natural correctness label, only downstream utility. Gains 3.3 points on CL-Bench and 6.7 on tau2-bench over the strongest baseline with a GPT-4o worker. Code is public. Transferable idea for anyone maintaining CLAUDE.md or SKILL.md files: A/B each edit against real task replays instead of accepting agent-proposed doc changes because they read well.