Automate iterative skill refinement through evaluation, diagnosis, and modification cycles.
npx skills add https://github.com/tya5/reyn --skill skill-improver-tya5
Please help me install this Agent Skill. Skill: skill_improver Source: https://github.com/tya5/reyn/tree/main/src/reyn/stdlib/skills/skill_improver Command: npx skills add https://github.com/tya5/reyn --skill skill-improver-tya5
This Skill automates the process of refining existing skills by managing iterative improvements and evaluations, reducing manual effort and ensuring systematic enhancement.
eval, analyzes scores, and plans minimal DSL changes to meet quality thresholds.Use the skill_improver to auto-correct and optimize your current skill based on evaluation results, aiming for higher scores in fewer iterations.
High-intent search queries and answers about installing and using this skill.
Skill refinement is automated by copying, evaluating, diagnosing, and modifying skill files in iterative cycles to achieve desired quality metrics. The process runs evaluation loops, analyzes scores, and plans minimal DSL changes to meet quality thresholds.
Automated skill evaluation works by running evaluations, analyzing resulting scores, and diagnosing deficits to plan minimal DSL changes. This loop repeats systematically until the skill achieves the desired quality threshold.
Improving prompt-answering skills to meet high evaluation scores involves continually adjusting prompts and logic through automated evaluation loops. The system targets higher scores in fewer iterations until exceeding the desired threshold, such as 0.9.
Yes, refined skill improvements can be copied back to the original skill repository. Once the iterative evaluation cycles achieve the desired quality metrics, the system automates copying the updated files back to your source.
To start automating skill improvements, you need an existing skill repository with evaluation capabilities. The system uses these to run evaluation loops, diagnose scores, and apply minimal DSL changes to meet quality thresholds.
An evaluation score plateau indicates the current DSL changes are insufficient. The system addresses this by diagnosing the specific deficits and planning targeted, minimal modifications to push the score past the desired threshold.
kzahediRun performance benchmarks on YARS simulations and detect regressions.
Ronin15Compare SDL3 HammerEngine metrics against baselines to flag performance regressions.
CommonCapitalDesign, test, and refine AI skills with evaluation feedback loops.
martin-hsu-testOrchestrate CEO, design, engineering, and DX reviews from disk into an auditable plan.