Skill
PulseAugur coverage of Skill — every cluster mentioning Skill across labs, papers, and developer communities, ranked by signal.
3 day(s) with sentiment data
-
AI Tooling: When to Use MCP vs. Skills
The article discusses the appropriate use of AI tools, differentiating between MCP (Model-Centric Programming) and skills. MCP is recommended for tasks requiring predictability, while skills are better suited for more c…
-
Ant Digital launches Agentar Financial Edition for AI agent deployment
Ant Digital has launched Agentar Financial Edition, a platform designed to help financial institutions build and manage AI agents. This new offering provides pre-built financial agent teams, industry-specific skills, an…
-
Indextkn API tracks pricing for 900 AI models
Indextkn is a new API service designed to track and provide real-time pricing for over 900 AI models. Developed to address the challenge of managing and comparing AI model costs, Indextkn fetches pricing data frequently…
-
Expert-informed skill prompting improves cross-dataset stability for Chinese metaphor identification
Researchers investigated methods for improving Chinese metaphor identification across different datasets. They compared four approaches: BERT fine-tuning (BERT-FT), QLoRA-based LLM fine-tuning (LLM-FT), direct zero-shot…
-
Claude Code users can now track token costs by Skill with a new script
A developer has created a shell script called `usage-breakdown.sh` to analyze token usage in Claude Code, as the built-in `/usage` command only provides per-model cost totals. The script parses `transcript.jsonl` files,…
-
New benchmark LongWoF-Bench evaluates reusable AI experience
Researchers have introduced LongWoF-Bench, a benchmark designed to evaluate the reuse of verified execution experience for large language models. This approach, termed EvoMap, consolidates successful task trajectories i…
-
Alipay's AI assistant "Xiaozheng" streamlines marriage registration services
Alipay's AI assistant, "Xiaozheng," has launched a new service for marriage registration, leveraging its MCP capabilities to connect with official service channels in ten provinces. This initiative aims to streamline th…
-
New LLM agent SKILL optimizes logic synthesis with multi-model approach
Researchers have developed SKILL, a novel agent that uses multiple large language models and reinforcement learning to optimize logic synthesis. The system employs GPT-4o for strategic planning, Claude Sonnet 4 for deta…
-
Cursor Learn Mode open-sourced to teach desktop workflows
An open-source tool called Cursor Learn Mode has been released, designed to teach desktop workflows by recording user actions. This tool, compatible with both Windows and macOS, captures demonstrated processes and saves…
-
HarmonyOS streamlines development with AI-powered 'service kitchen'
HarmonyOS is enhancing its developer experience with a new "service kitchen" approach, leveraging the Atomic Service Cross Framework (ASCF) to allow developers to reuse existing code from other platforms like mini-progr…
-
New framework improves AI compliance detection for long SKILL documents
Researchers have developed SkillCDG, a novel graph-based framework designed to improve the efficiency and accuracy of compliance detection for long SKILL documents in enterprise agent systems. This framework represents …
-
Enterprise AI adoption: Leveraging DESIGN.md and SKILL for design systems
This article discusses how to leverage DESIGN.md and SKILL within an enterprise setting to enhance the impact of design systems in the era of AI. It aims to share practical implementation strategies and lessons learned …
-
GPT 5.6 Luna reportedly available for free ChatGPT users with new Agent Plugins
A new AI development, GPT 5.6 Luna, is reportedly available without restrictions for free users of the web version of ChatGPT. This advancement includes the introduction of Agent Plugins, which combine Skill and MCP fun…
-
Claude Code skills face description limits, impacting AI's ability to select tools
Claude Code, a feature within Anthropic's Claude AI, has limitations on how much descriptive text it can process for its skills. Two main constraints are at play: a per-skill limit of 1,536 characters for descriptions a…
-
AI enhances chess gameplay through DeltaChat and Webxdc integration
The user is exploring the integration of AI into their chess game, specifically using tools like DeltaChat and Webxdc. They are working on improving their chess skills by reverse-engineering chatmail and client function…
-
AI agents struggle with tool noise, increasing costs and confusion
The article discusses the problem of context bloat in AI agent harnesses, where excessive tool information overwhelms language models and increases token costs. It highlights how tools like Google Calendar's events.list…
-
Transit agent experimenter rolls back SKILL layer after MCP tool failure
An experimenter encountered issues with a transit agent built on the MCP framework. After introducing a SKILL layer to manage MCP tool usage, the agent began to favor web search over the MCP tool, leading to inaccurate …
-
Apidog enhances API development with new CLI and SKILL for complex workflows
Apidog has developed a new Command Line Interface (CLI) and SKILL system to complement its existing MCP tool, aiming to handle more complex API development workflows. While MCP excels at standardized tool connections fo…
-
AI agents can be more scalable and maintainable using reusable 'Skills'
This article proposes a new approach to building AI agents by packaging their behavior into reusable 'Skills' rather than relying on large, monolithic prompts. A Skill encapsulates specific functionalities, such as revi…
-
Microsoft report: Employees ready for AI, but companies lag behind · 2 sources tracked
A recent report from Microsoft indicates that while employees are ready to embrace AI in the workplace, many organizations and leaders are lagging behind. The report, based on a survey of 20,000 AI users and analysis of…