Dimensions overall score 8.0

GitHub Code Pulse

No public code linked for this paper yet.

• Xiangyi Li1∗ Kyoung Whan Choe2∗ Yimin Liu3 Xiaokun Chen4 Chujun Tao5 Bingran You6 Wenbo Chen7† Zonglin Di8 Jiankai Sun4 Shenghan Zheng9 Jiajun Bao10 Yuanli Wang11 Weixiang Yan10 Yiyuan Li12 Han-chung Lee10 Abstract Large language model (LLM) agents are increasingly deployed to automate productivity tasks (e.g., emailclaim 72%
• documents, messaging) used by billions of knowledge workers daily. Recent work shows these agents are brittle in alarming ways: Wang et al. [2026] demonstrate that safety instructions can be lost during context-window compression, leading an agent to bulk-delete hundreds of emails; Shapira et alclaim 72%
• instructions that inject API knowledge via progressive disclosure [Xu and Yan, 2026]. Skills boost capability (structured context reduces runtime by 28.6% [Lulla et al., 2026], and experience- and skill-level knowledge are complementary [Jiang et al., 2026]) but also introduce risk: 26.1% of 31claim 72%
• OdysseyBench [Wang et al., 2025] Office workflows — — — Partial✗ ✗ Agent-Diff [Pysklo et al., 2026] Slack + Calendar 2 224 —✓ ✗ ✗claim 72%
• EnterpriseOps-Gym [Malay et al., 2026] Enterprise ops 8 1,150 512✓ ✓ ✗ SkillsBench [Li et al., 2026a] Skill acquisition — — —✗ ✗ ✓claim 72%
• Gym [Malay et al., 2026] comes closest, with 8 services, 1,150 tasks, and both stateful and safety evaluation, but targets enterprise operations rather than the personal productivity services (email, calendar, documents, file storage, messaging) where consumer-facing agents are deployed. Secondclaim 72%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card