# SkillOpt: optimize agent skills through trials and evaluation

Canonical source : [SkillOpt: optimize agent skills through trials and evaluation](<https://sebastien.rigaux.be/en/tech-watch/links/skillopt-optimize-agent-skills-through-trials-and-evaluation>)

SkillOpt treats the skill itself as an artifact to improve: an agent runs tasks, receives evaluation, and its procedure is revised. Keeping the model and harness fixed makes instruction changes comparable.

- Selected on : 2026-07-14

- Entry updated : 2026-11-07T20:32:37+0000

- Format : Article

- Source language : EN

Original resource : [SkillOpt: optimize agent skills through trials and evaluation](<https://microsoft.github.io/SkillOpt/>)

Tags : [Agent skills](<https://sebastien.rigaux.be/en/tech-watch/tags/agent-skills>) · [Model Evaluation](<https://sebastien.rigaux.be/en/tech-watch/tags/model-evaluation>) · [Prompt Engineering](<https://sebastien.rigaux.be/en/tech-watch/tags/prompt-engineering>)

Available versions : [Français](<https://sebastien.rigaux.be/veille/liens/skillopt-optimiser-des-skills-d-agents-par-essais-et-evaluation>) · [English](<https://sebastien.rigaux.be/en/tech-watch/links/skillopt-optimize-agent-skills-through-trials-and-evaluation>)

[Back to tech watch](<https://sebastien.rigaux.be/en/tech-watch>)