On Skill (AI) is a package of expertise and instructions that shows an AI agent how to perform a specific task. Skills expand an agent's capabilities by providing specialized knowledge and repeatable workflows.
Building on the open format of Agent Skills
Anthropic developed the Agent Skills format and published it as an open standard. In this format, a skill is a folder containing a SKILL.md file. The file contains metadata (at least a name and a description) and the actual instructions. The description specifies what the skill does and when the agent uses it. Optionally, a skill bundles scripts, reference documents, templates, and data files.
Functionality
The agent loads skills in three stages. Initially, at startup, it only reads the name and description of each skill. If a task matches the description, it then reads the complete instructions. After that, it executes the instructions and, if necessary, downloads scripts or files. Therefore, an agent has many skills ready and takes up little space in the context.
Business significance
Skills packages process knowledge as well as company-, team-, and user-specific context into versioned folders. This allows companies to store specialized knowledge as reusable instructions. At the same time, they turn multistep tasks into uniform, verifiable workflows. A skill can be used by any agent that supports the format. This includes, among others, Claude, GitHub Copilot, Cursor, VS Code, ChatGPT, and Codex, as well as Gemini CLI.
Example SAP Business One
An example from the SAP Business One environment is the Versino Skills: curated knowledge packages covering exactly one SAP B1 topic each.
Demarcation
A skill provides knowledge and instructions. Tool usage, on the other hand, means that an AI model integrates external tools operationally, for example to execute program code. However, a skill can also bring scripts along, and the experimental field allowed-tools lists pre-approved tools. RAG retrieves information from a database or knowledge base. A skill, on the other hand, gives the agent instructions for a specific task.
Related entries: Versino Skills; AI assistant; AI agent for SAP B1; RAG (Retrieval-Augmented Generation) in ERP context; Large Language Model (LLM); Model Context Protocol (MCP).
AI assistant meets SAP Business One: Two paths out of the Versino AI shop
Sprint and Swirl: How an AI agent divides work and learns from trials
LATS: Considering several paths simultaneously with AI — and the availability check in SAP B1
ReAct and RLEF: How Artificial Intelligence Learns from Real Mistakes — and What Automated Bank Reconciliation in SAP Business One Has to Do with It
SAP B1 10.0 FP2608: Service Layer AI as a transactional layer
