📊 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.
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Anthropic has shared a new approach to AI skills, framing them as folders containing instructions, scripts, and assets rather than simple prompts. This method enhances consistency, onboarding, and continuous improvement in AI workflows. The company ran hundreds of experiments to develop this model, emphasizing the importance of structured, reusable units for organizational AI deployment.
Anthropic has revealed that its approach to building AI capabilities involves creating ‘Skills’ as structured folders rather than simple prompts, a shift that could significantly change organizational AI workflows. This insight, shared in a recent internal publication, underscores a move towards durable, reusable units that encapsulate instructions, code, and reference data. The development aims to improve consistency, onboarding, and continuous refinement of AI agents within companies, making their deployment more reliable and scalable.
According to an account from a Claude Code engineer, Anthropic’s ‘Skills’ are fundamentally redefined as folders containing instructions, scripts, templates, reference documents, and configuration data, rather than just saved prompts. This structural change allows AI agents to discover, read, and execute complex sets of instructions dynamically, fostering more robust and maintainable workflows.
Anthropic’s internal research indicates that these Skills serve three core functions for organizations: ensuring consistent output regardless of the user, compressing onboarding by encapsulating tribal knowledge, and enabling continuous improvement through iterative refinement. The company emphasizes that a well-developed Skill can evolve over time as edge cases are encountered and documented, transforming it into an asset that appreciates in value.
Anthropic identified nine categories of Skills, ranging from API references and data analysis to operational runbooks and infrastructure management. The most valuable category, according to the company, is verification — the Skills that check and validate outputs, which directly impact output quality and reduce errors.
A Skill is a folder, not a prompt
Anthropic published what it learned running hundreds of Skills across its own engineering org. Read as a business memo, the point is bigger than a coding trick: this is how ad-hoc prompting becomes durable institutional capability — the SOPs your agents actually follow, versioned and shared.
“A Skill is just a clever markdown prompt you save in a file.”
A folder the agent can discover, read & run — instructions, scripts, references, templates, config & on-demand hooks.
The knowledge of how your organization actually operates can be captured, versioned, shared & executed — and the thing capturing it is a humble folder with a script and a gotchas list inside. For the builder, that’s context engineering with real tools attached. For whoever owns the budget, it’s the difference between AI that starts from zero every morning and an asset that compounds. Caveats: best practices are still evolving, checked-in Skills cost context, and curation beats accumulation. Start with one Skill, one gotcha, and the category that catches your mistakes.
Implications for Organizational AI Deployment
This development signifies a shift from ephemeral prompt engineering to durable, asset-based AI capabilities. By structuring Skills as folders, organizations can create reusable, versioned, and sharable units that improve consistency, reduce onboarding time, and facilitate continuous improvement. This approach could standardize how companies build, maintain, and evolve their AI workflows, leading to more reliable and scalable AI systems.
Furthermore, viewing Skills as assets rather than prompts aligns with broader trends in enterprise AI, where institutional memory, guardrails, and reusable components are critical for operational success. The emphasis on verification Skills highlights the importance of quality control in automated workflows, potentially reducing errors and increasing trust in AI outputs.
Products Worth Considering
DJI Osmo 360 Road Cycling First-Person Combo 360° Action Camera
Samsung 49-Inch Odyssey G9 G91F Curved Dual QHD Gaming Monitor, 144Hz
Samsung Odyssey G93SC 49-Inch QD-OLED Curved Gaming Monitor
As an affiliate, we earn on qualifying purchases.
From Prompting to Asset Building in AI Engineering
Prior to this development, most organizations relied heavily on prompt engineering—crafting specific instructions each time an AI task was performed. This ad-hoc approach often led to inconsistent results and onboarding challenges. Anthropic’s internal research, shared publicly in a recent memo, demonstrates a move towards creating structured, reusable units that encapsulate organizational knowledge and operational procedures.
Historically, AI development has focused on prompt tuning and fine-tuning models. Anthropic’s approach emphasizes building comprehensive ‘Skills’ that serve as containers for instructions, scripts, and reference data, effectively turning AI capabilities into organizational assets. This methodology echoes best practices in software engineering, where modular, versioned components enhance maintainability and scalability.
The company’s experiments with hundreds of Skills across nine categories suggest that this model not only improves performance but also enables continuous refinement, turning initial efforts into valuable, evolving assets.
“A Skill is not just a prompt saved in a file; it’s a folder containing instructions, scripts, and reference data that the agent can discover and execute.”
— Thorsten Meyer, AI researcher at Anthropic
Unresolved Questions About Skills Implementation
It is not yet clear how widely this folder-based Skills approach has been adopted across different organizations or how it performs at scale outside Anthropic. Details on integration challenges, tooling support, and the impact on existing workflows remain limited. Additionally, how this approach compares in cost and complexity to traditional prompt engineering is still under evaluation.
Next Steps for Broader Adoption and Validation
Organizations interested in this approach will likely begin experimenting with creating their own Skills as structured folders, focusing on categories like verification and automation. Further research and case studies are expected to emerge, evaluating the scalability and effectiveness of this method. Anthropic may also release tooling to facilitate the development and management of Skills, making this approach more accessible to a wider audience.
Key Questions
How does a Skill differ from a traditional prompt?
A Skill is a structured folder containing instructions, scripts, templates, and reference data, whereas a prompt is typically a single text instruction. Skills enable reusable, versioned, and more complex operational units for AI agents.
Why are Skills considered more durable assets?
Because they encapsulate organizational knowledge, guardrails, and procedures in a structured format that can evolve over time, making them valuable assets rather than ephemeral prompts.
What is the most valuable category of Skills according to Anthropic?
Verification Skills, which check and validate outputs, are considered the most impactful because they directly improve output quality and reduce errors.
Will this approach replace prompt engineering entirely?
It is too early to say, but this structured, asset-based approach aims to complement prompt engineering by providing more reliable, maintainable, and scalable capabilities for organizations.
What are the main challenges in adopting Skills as folders?
Potential challenges include tooling support, integration with existing workflows, and managing version control at scale. Further testing is needed to evaluate these factors fully.
Source: ThorstenMeyerAI.com
