Fetching from the wire…
Research2026-05-25 · source-backed
SkillOpt proposes treating agent skills as external state of a frozen LLM, optimized with the same discipline as weight-space training. Rather than hand-crafting or loosely self-revising skills, SkillOpt applies controllable text-space optimization that reliably improves performance. First framework to bring reproducible optimization to the skill layer of agent systems.
Each link below shares sources, entities, or timing with this story.
Codex supports Agent Skills / Shared entities / Same source / Shared topic / What happened next
Linked by a graph relationship (Codex supports Agent Skills); both cover LLM, SkillOpt; cite the same source (SkillOpt).
Codex supports Agent Skills / Shared entity: LLM / Same source domain / Shared topic / What happened next
Linked by a graph relationship (Codex supports Agent Skills); both cover LLM; reported by the same outlet (arxiv.org).
Linked by a graph relationship (Codex supports Agent Skills); both cover LLM; reported by the same outlet (arxiv.org).
Vercel supports Agent Skills / Shared entity: LLM / Shared topic / Earlier coverage
Linked by a graph relationship (Vercel supports Agent Skills); both cover LLM; overlapping topics (agent, external, framework, layer).
Codex supports Agent Skills / Shared entity: LLM / Same source domain / Shared topic / What happened next
Linked by a graph relationship (Codex supports Agent Skills); both cover LLM; reported by the same outlet (arxiv.org).
Microsoft released SkillOpt / Shared entity: LLM / Same source domain / What happened next / Tension
Linked by a graph relationship (Microsoft released SkillOpt); both cover LLM; reported by the same outlet (arxiv.org).
OpenAI supports Agent Skills / Shared entity: LLM / Same source domain / Shared topic / What happened next
Linked by a graph relationship (OpenAI supports Agent Skills); both cover LLM; reported by the same outlet (arxiv.org).
Microsoft released SkillOpt / Shared entity: Agent Skills / Shared topic / What happened next
Linked by a graph relationship (Microsoft released SkillOpt); both cover Agent Skills; overlapping topics (agent, framework, skill).