📊 Full opportunity report: A Skill Is a Folder, Not a Prompt: What Anthropic Learned Running Hundreds of Them on ThorstenMeyerAI.com — validation score, market gap, and execution plan.

Anthropic has announced a new approach to AI agent management, defining ‘Skills’ as comprehensive folders that contain instructions, scripts, and reference assets, rather than simple prompts. This shift aims to make AI outputs more consistent, facilitate onboarding, and build a durable institutional knowledge base. The development is based on Anthropic’s experience running hundreds of Skills internally, offering a new model for organizations deploying AI at scale.

In a recent publication, Anthropic detailed its findings from deploying hundreds of Skills across its engineering teams. Unlike traditional prompt engineering, where instructions are often single text prompts, a Skill is now understood as a folder that can include instructions, reference documents, runnable scripts, templates, data, configurations, and even hooks that activate during operation. This change redefines how organizations can structure their AI workflows, making them more modular, reusable, and maintainable.

Anthropic emphasizes that Skills serve three core functions: ensuring consistent output regardless of who runs the agent, compressing onboarding processes by encapsulating tribal knowledge, and enabling continuous improvement through iteration. The company reports that its best Skills began as simple caveats but became more refined, accumulating edge cases and best practices over time. The result is a library of Skills that act as an organizational asset, not just a prompt or note.

Anthropic identified nine categories of Skills, ranging from library references and product verification to code scaffolding and infrastructure operations. The most impactful, according to the company, is verification Skills—those that check and validate outputs—because they directly improve output quality and reduce errors. The approach encourages organizations to focus on building robust, specific Skills that push the model off its defaults and capture institutional knowledge.

Transforming AI Workflow Management in Organizations

This development matters because it shifts the way organizations think about deploying AI agents. Moving from ad-hoc prompts to structured Skills as folders creates a durable, scalable framework for operational consistency, knowledge retention, and continuous improvement. It reduces reliance on individual expertise and helps embed best practices into the AI system itself, potentially lowering costs and increasing reliability over time.

For businesses, adopting this approach could mean more predictable AI outputs, faster onboarding of new team members, and a strategic asset that grows more valuable as it accumulates institutional knowledge. It also signals a move toward more formalized, versioned, and sharable AI assets within organizations, akin to software libraries or operational playbooks.

From Prompt Engineering to Asset Building

Prior to this shift, most AI teams relied heavily on prompt engineering—crafting specific prompts to guide model behavior. These prompts were often one-off and lacked durability, requiring repeated re-creation or adjustment. Anthropic’s internal experience revealed that prompts alone are insufficient for scaling AI deployment in complex workflows.

By contrast, the concept of Skills as folders encapsulates knowledge, procedures, and code into a single, versioned asset. This approach aligns with broader trends in software engineering, emphasizing modularity, reusability, and institutional memory. Anthropic’s insights are based on running hundreds of Skills internally, which demonstrated the value of structured, comprehensive units for operational consistency and continuous improvement.

This development builds on prior work in AI prompt engineering but represents a significant evolution toward more durable, organizationally embedded AI capabilities.

“A Skill is not just a prompt saved in a file; it’s a container that includes instructions, scripts, and assets, making it a true organizational asset.”

— Thorsten Meyer, AI researcher

What Aspects of Skills Are Still Unclear?

It is not yet clear how widely this approach will be adopted outside Anthropic or how it will scale in different organizational contexts. Details about the specific technical implementation, integration challenges, and long-term maintenance are still emerging. Additionally, the precise impact on productivity and error rates across diverse teams remains to be quantified.

Future Steps for Organizational AI Asset Management

Organizations interested in this approach should begin cataloging their internal procedures and knowledge into structured folders, testing how these Skills improve consistency and onboarding. Further research and case studies are expected to evaluate the scalability and long-term benefits of this model. Anthropic itself is likely to refine its Skills library and share best practices for broader adoption.

Key Questions

How does defining a Skill as a folder improve AI deployment?

It creates a durable, reusable asset that encapsulates instructions, scripts, and knowledge, leading to more consistent outputs, easier onboarding, and continuous improvement.

What are the main categories of Skills identified by Anthropic?

They include library references, product verification, data analysis, business automation, code scaffolding, review, deployment, runbooks, and infrastructure operations.

Can this approach be applied outside of AI development teams?

Yes, any organization seeking to formalize and institutionalize procedural knowledge could benefit from structuring it into Skills folders, though implementation details may vary.

What challenges might organizations face adopting this model?

Potential challenges include technical integration, maintaining version control, ensuring proper documentation, and training staff to develop and update Skills effectively.

Source: ThorstenMeyerAI.com