Agent Learning Daily Digest — 2026-05-15

数据来源:GitHub API(6 组关键词)+ HN RSS + arXiv
原始数据:2026-05-15
今日采集 140 条,筛选 11 条高信号内容

今日高信号

1. 🔐 Toward Securing AI Agents Like Operating Systems (arXiv)

2. 🧪 Holistic Evaluation and Failure Diagnosis of AI Agents (arXiv)

3. 🏗️ claude-code-harness — Plan→Work→Review 自主循环 (⭐866)

4. 📊 CodexBar — Codex & Claude Code 用量统计 (⭐12191)

5. 🤖 DeepSeek-Reasonix — DeepSeek 原生 coding agent (⭐2440)

6. 🎮 BitFun — 桌面 Agent 运行时 (⭐728)

7. 📋 Systematically Auditing AI Agent Benchmarks with BenchJack (arXiv)

8. 🛡️ mcp-firewall — MCP 安全策略执行 (⭐1)

9. 🔬 AI Harness Engineering — Agent 运行时基底理论 (arXiv)

10. 🧬 Skill-R1: Agent Skill Evolution via RL (arXiv)

11. 📜 Making OpenAPI Documentation Agent-Ready (arXiv)


观察清单