Weekly Trending
Brainhuggers BureauGitHub Trending
01→
VoltAgent awesome-design-md
+1.2k this wk · 99.5k starsDESIGN.md turns brand systems into executable instructions for coding agents
VoltAgent's awesome-design-md collects DESIGN.md files from popular brand design systems and frames them as reusable instructions for coding agents that generate matching UI.
02→
iOfficeAI OfficeCLI
+802 this wk · 9.8k starsiOfficeAI/OfficeCLI
OfficeCLI exposes Word, Excel, and PowerPoint documents through an agent-native command line so document work can become scripted, inspected, and automated.
03→
TencentDB Agent Memory
+567 this wk · 7.2k starsTencentCloud/TencentDB-Agent-Memory
TencentDB Agent Memory packages agent memory as a tiered local infrastructure component, turning recall into a concrete system layer rather than a vague assistant promise.
04→
bradautomates/claude-video
+539 this wk · 4.2k starsbradautomates/claude-video
A Claude Code skill that downloads videos, extracts frames, transcribes audio, and packages video into agent-readable context.
05→
OthmanAdi/planning-with-files
File-based planning becomes durable agent infrastructure
OthmanAdi/planning-with-files packages persistent markdown plans, ledgers, completion gates, and shared state as infrastructure for AI coding agents and long-running tasks.
HF Papers
01→
Securing the AI Agent
Securing the AI Agent: A Unified Framework for Multi-Layer Agent Red Teaming
AI-Infra-Guard frames agent security as a layered attack surface across infrastructure, MCP/tool use, skills, behavior, and model-level jailbreaks.
02→
SkillOpt-Lite
SkillOpt-Lite: Better and Faster Agent Self-evolution via One Line of Vibe
SkillOpt-Lite improves agent skills through trajectory exploration, consensus attribute mining, and validation gates, turning prior execution traces into better reusable procedures.
03→
AgentLens
AgentLens makes coding-agent evaluation a trajectory review
AgentLens evaluates coding agents by reviewing the full production trajectory, including tool use, recovery, verification, and communication, instead of judging only the final answer.
04→
Remember When It Matters
Remember When It Matters: Proactive Memory Agent for Long-Horizon Agents
A separate memory agent monitors long-horizon work and injects reminders only when behavior-relevant state is about to decay.
05→
UniClawBench
UniClawBench: A Universal Benchmark for Proactive Agents on Real-World Tasks
A proactive-agent benchmark built around real-world tool use, exploration, long context, multimodal understanding, and cross-platform coordination in live Docker environments.
06→
Ideas Have Genomes
Ideas Have Genomes: Benchmarking Scientific Lineage Reasoning and Lineage-Grounded Idea Generation
A benchmark that represents scientific ideas as evidence-grounded genome objects, tracking inheritance, mutation, loss, import, and novelty across lineages.