Tag
本文揭示了自我改进智能体中技能-策略共进化的脆弱性陷阱，提出部署导向的评估协议CODA来审计技能漂移的影响，发现最差情况下共进化增益会被削弱52-93%。
A curated accounting skill library for AI finance agents covering close, reporting, reconciliation, AP, and AR workflows, built by analyzing prompts across finance-specific skill libraries.
SkillForge proposes an agentic reinforcement learning method that co-evolves an LLM agent and its skill library through a fitness-driven lifecycle of trial, active, stable, and retired states. It reports up to 7.8% relative improvement over strong baselines on interactive agent benchmarks and releases SkillFurnace, a 5k+ record dataset for studying skill quality and lifecycle management.
作者发布了一套 AI agent 技能（skills），旨在对抗「脑腐」（brain rot）——即在使用 AI agent 构建时把思考外包给 AI、导致产出变成 slop 的问题。这些技能帮助用户在通用构建任务、UI/UX 打磨与算法能力训练中主动思考。
A tweet discusses the missing piece in self-improving agents: a strict review gate where an experienced engineer must approve each skill. It explains how a growing library improves discovery and extraction quality, with humans approving all merges to keep the system trustworthy.
Microsoft AI Frontiers introduces Web Skill Factory, a pipeline that converts solved web tasks into reusable, verified code programs. Reusing the library on a WebArena subset with gpt-5.4 boosts held-out accuracy from 55% to 70% and reduces average steps from 17.1 to 14.7.
Introduces SkillCenter, the largest open skill library for AI agents with 216,938 structured, source-grounded skills across 24 domains, built using an automated pipeline with an LLM-based quality gate called SkillGate. The library combines peer-reviewed and community skills, emphasizing traceability and offline searchability.
Fable 5 is transitioning to a pay-per-use model; users are advised to have it train its replacement and write skill files for continued use. The thread details a prompt to have Fable audit a repo and create a comprehensive skill library.
NVIDIA's ASPIRE framework enables robots to build a persistent library of skills from successful experiences, allowing reuse for new tasks and improving learning efficiency over time.
Introducing ASPIRE, a framework for robots to continuously evolve a library of skills through evolutionary search and distillation, enabling efficient sim-to-real and cross-embodiment transfer with up to 10x reduction in transfer learning tokens. The full stack is open-sourced.
VisualSkill proposes a hierarchical multimodal skill library for computer-use agents that combines text and figures, achieving a 15.3 point absolute lift on CUA benchmarks over text-only baselines by retaining visual information for GUI interaction.
Introduces Playful Agentic Robot Learning, where embodied coding agents use self-directed play to learn reusable skills, improving downstream task performance without additional training. The proposed RATs system achieves significant gains over baselines in simulation and real-world transfer.
This paper proposes Embodied-BenchClaw, an autonomous multi-agent system that automatically constructs embodied spatial intelligence benchmarks from user intent through a five-stage pipeline with process quality control and an extensible Skill Library.
This article argues that the first AI strategy for companies should be creating a 'skill library' to capture the reusable working methods of top performers, so agents can learn the method behind tasks rather than just access data. It promotes a live webinar called Skills 101.
This MUSE-Autoskill paper focuses on how an Agent manages an entire skill library, placing skills into a complete lifecycle: creation, memory, management, evaluation, and re-optimization.
SkillGraph is a framework that represents reusable skills as nodes in a directed graph to enable large language model agents to handle compositional tasks more effectively through structured skill retrieval and continuous evolution.
Skill1 is a unified framework that trains a single policy to co-evolve skill selection, utilization, and distillation using a shared task-outcome objective. Experiments on ALFWorld and WebShop show it outperforms existing baselines in complex task environments.
AI Design Skills 是一个开源的 AI 设计技能合集，为 Claude Code、Cursor、Codex、Windsurf 等 AI 编码工具提供自包含的 Markdown 规则文件，目前包含 landing-page-design 技能，可通过复制 SKILL.md 轻松安装和自定义。
Agentic Awesome Skills (AAS Core) is an open-source library of 2,724+ installable SKILL.md playbooks that lets agents like Codex and Claude search a local skill catalog, record selections, and preview plans before applying changes. The project supports a wide range of AI coding tools including Claude Code, Cursor, Gemini CLI, and Copilot.