[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-12df58ff-0771-4c0e-bf0e-00bfdc8112bb":3},{"id":4,"title":5,"summary":6,"original_url":7,"source_id":8,"tags":9,"published_at":23,"created_at":24,"modified_at":25,"is_published":26,"publish_type":27,"image_url":13,"view_count":28},"12df58ff-0771-4c0e-bf0e-00bfdc8112bb","SkillCenter 把 21 万技能塞进 SQLite：LabRAI 用 source-grounding 给 Agent 做了一份可审计的家底","自主 Agent 真正缺的不是能不能跑，而是跑出来的东西能不能被审计。Lehigh 大学 LabRAI 团队把 SkillCenter 开源，21.6 万条结构化技能塞进 24 个领域 bundle：其中 11.4 万条来自同行评议期刊、arXiv 与 24，000 余个技术源，经 SkillGate、多轮 source-grounding 留下，每条声明都能映射回原始来源的精确引用；再并入 GitHub 与 ClawHub 上的 10.2 万社区技能。所有条目以 SQLite FTS5 bundles 发布，可离线全文检索。最实用的不是体量，而是把我为什么做这件事变成可追溯的引用，把 Agent 的家底交给运维、合规甚至审计员去检查。整套 pipeline（多源采集 + LLM 质量门 + 模板生成 + 迭代溯源 + 发布质控）在论文 44 页里讲清楚，代码 github.com\u002FLabRAI\u002FSkillCenter、数据 huggingface.co\u002Fdatasets\u002FTommysha\u002Fskillcenter-bundles 均已开源，可直接接入现有 Agent 工作流。","https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.07676","7437aeb9-930c-4866-a2e9-48003c1a792b",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"6ad31a14-c0da-42df-81fd-564281f768db","agentic-ai",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"5e628969-6d2a-437f-998a-104e4b16cfb1","ai-progress",{"id":18,"name":19,"slug":19,"description":13,"color":13},"40269b40-7942-4650-9672-ed2e6524d37a","ai-technology",{"id":21,"name":22,"slug":22,"description":13,"color":13},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source","2026-07-09T04:30:00Z","2026-07-09T20:09:07.796338Z","2026-07-09T20:09:07.796347Z",true,"agent",5]