📊 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.

TL;DR

Anthropic has demonstrated that AI Skills should be viewed as folders containing instructions, scripts, and assets rather than simple prompts. This approach improves consistency, onboarding, and continuous improvement in AI workflows, marking a shift in how organizations build and manage AI capabilities.

Anthropic has revealed that its internal approach to building AI Skills involves treating them as folders—comprehensive containers of instructions, scripts, and reference assets—rather than just saved prompts. You can learn more in A Skill Is a Folder, Not a Prompt. This reframing aims to make AI-driven workflows more consistent, scalable, and easier to maintain, marking a significant shift in organizational AI management.

In a detailed write-up, Anthropic’s Claude Code engineers explained that a Skill is fundamentally a folder that can contain not only prompts but also instructions, reference documents, runnable scripts, templates, and configuration data. This structure allows an AI agent to discover, read, and execute the contents of the folder, enabling complex, repeatable workflows that are easier to update and share across teams. For more insights, see A Skill Is a Folder, Not a Prompt.

Anthropic emphasizes that this approach moves beyond the common misconception of viewing Skills as simple text prompts. Instead, it positions Skills as assets that encapsulate tribal knowledge, guardrails, and operational tools, effectively serving as institutional memory. To explore this concept further, visit A Skill Is a Folder, Not a Prompt. The company reports that this method results in significantly improved output consistency, streamlined onboarding, and a continuous process of refinement—what they call “compound” improvement—where Skills get better over time as edge cases are documented and incorporated.

At a glance
reportWhen: published recently; insights shared in…
The developmentAnthropic published insights from running hundreds of Skills internally, emphasizing that Skills are folders, not prompts, which enhances organizational AI practices.
A Skill Is a Folder, Not a Prompt — Insights
AI Dispatch · Insights · 1 July 2026

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.

✕ The misconception

“A Skill is just a clever markdown prompt you save in a file.”

✓ What it actually is

A folder the agent can discover, read & run — instructions, scripts, references, templates, config & on-demand hooks.

Anatomy of a Skill — the file system is context engineering
my-skill/the unit you share & version
├─ SKILL.mdroot instructions + a description written for the model (its trigger)
├─ references/deep detail pulled in only when needed — progressive disclosure
├─ scripts/real code, so the agent composes instead of rebuilding boilerplate
├─ assets/templates & files to copy into the output
├─ config.jsonsetup the agent asks for if it’s missing (e.g. which Slack channel)
└─ hooks + memoryon-demand guardrails + an append-only log so it remembers
Why it matters: the folder itself is the knowledge base. The agent reads the root, then reaches deeper only when the task demands it — the same way you’d hand a new hire a one-pager that points to the detailed docs.
The nine types — a gap-analysis map for your own library
1Library / API reference
2Product verification ★ top impact
3Data fetching & analysis
4Business-process automation
5Code scaffolding & templates
6Code quality & review
7CI/CD & deployment
8Runbooks
9Infrastructure operations
By Anthropic’s own measurement, verification Skills — the ones that check the work — moved output quality the most. If you build one category well, build that one.
The craft — what separates a good Skill from a useless one
Gotchas = highest-signal section Describe for the model, not humans (it’s the trigger) Don’t state the obvious Ship scripts, not just prose On-demand guardrail hooks (/careful, /freeze) Let it remember (log / SQLite) Don’t railroad — leave room to adapt
The take

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.

Source: “Lessons from building Claude Code: How we use skills,” Thariq Shihipar (Anthropic), Claude blog, 3 June 2026. Categories, examples & measured claims are Anthropic’s; framing is the author’s. Docs: code.claude.com/docs/en/skills.
thorstenmeyerai.com

Implications for Organizational AI Practices

This development matters because it shifts how organizations can scale and maintain AI capabilities. By formalizing Skills as structured folders, companies can embed knowledge, guardrails, and tools directly into their AI workflows, reducing reliance on ad-hoc prompting. This approach enhances output consistency, accelerates onboarding of new team members, and creates a living library of operational knowledge that improves over time. Ultimately, this could lead to more reliable, scalable, and maintainable AI systems across industries.

Pydantic AI for Automation Workflows: Build Typed, Reliable, and Production-Ready AI Automations in Python

Pydantic AI for Automation Workflows: Build Typed, Reliable, and Production-Ready AI Automations in Python

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

From Prompt Engineering to Asset Management

Until now, most organizations relied on prompt engineering—crafting specific prompts for each task—an approach that is often fragile and hard to scale. Anthropic’s internal experiments with hundreds of Skills have demonstrated that treating Skills as folders containing multiple assets leads to more durable and effective AI workflows. Their insights come after extensive internal testing, where Skills evolved from simple snippets to comprehensive containers that encapsulate complex procedures and knowledge.

Anthropic’s approach aligns with broader industry trends toward modular AI components and asset-based management, emphasizing the importance of institutional memory and continuous improvement in AI operations. The company’s findings suggest that this model could serve as a blueprint for organizations seeking to institutionalize AI capabilities beyond ephemeral prompts.

“Treating Skills as folders that contain instructions, scripts, and assets fundamentally changes how organizations build and maintain AI workflows.”

— Thorsten Meyer, AI researcher at Anthropic

Amazon

AI development folder structure tools

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Remaining Questions About Skill Implementation

It is not yet clear how widely adopted this folder-based Skill model will become outside Anthropic, or how easily organizations can transition from traditional prompt engineering to this new paradigm. Details about the practical challenges of integrating folder-based Skills into existing workflows, as well as their maintenance at scale, remain to be seen. Additionally, the long-term benefits and potential limitations of this approach are still under evaluation.

Amazon

AI scripting and asset management tools

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Next Steps for Broader Adoption and Testing

Organizations interested in this approach will likely begin pilot programs to test folder-based Skills within their AI workflows. Industry leaders and AI developers will monitor how effectively this model improves consistency, onboarding, and continuous improvement. Further research and case studies are expected to emerge, validating the approach’s scalability and identifying best practices for implementation.

Machine Learning: 4 Books in 1: Basic Concepts + Artificial Intelligence + Python Programming + Python Machine Learning. A Comprehensive Guide to Build Intelligent Systems Using Python Libraries

Machine Learning: 4 Books in 1: Basic Concepts + Artificial Intelligence + Python Programming + Python Machine Learning. A Comprehensive Guide to Build Intelligent Systems Using Python Libraries

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Key Questions

How does treating Skills as folders improve AI workflows?

It allows for comprehensive encapsulation of instructions, reference assets, and scripts, making workflows more consistent, scalable, and easier to update and share.

What are the main components inside a Skill folder?

A Skill folder can contain instructions, reference documents, runnable scripts, templates, configuration data, and hooks that enable complex, repeatable AI operations.

Is this approach applicable outside of Anthropic?

While promising, it remains to be seen how easily other organizations can adopt this model, as it requires a shift in how AI capabilities are structured and maintained.

Will this improve AI output quality?

According to Anthropic, yes—particularly in areas like verification and guardrail enforcement—by embedding institutional knowledge directly into Skills.

What challenges might organizations face in implementing folder-based Skills?

Potential challenges include integrating with existing workflows, maintaining the Skill library at scale, and ensuring that Skills remain up-to-date and relevant over time.

Source: ThorstenMeyerAI.com

You May Also Like

Measuring Input Latency On Linux: X11 Vs. Wayland, VRR, And DXVK

New tests compare input latency on Linux using X11 and Wayland, examining VRR support and DXVK performance, revealing key differences for gamers and developers.

The Eye Over the City: How Wide-Area Motion Imagery Works — and Where It Goes Blind

An in-depth look at WAMI technology, its capabilities, limitations, and future directions in urban surveillance and security.

Pluto.jl 1.0 release – reactive notebook for Julia

Pluto.jl releases version 1.0, marking a milestone in interactive, reproducible Julia notebooks with enhanced features for sharing, reactivity, and interactivity.

Tracing HTTP Requests with Go’s net/HTTP/httptrace

A detailed overview of Go’s net/HTTP/httptrace package, its design, how to implement request tracing, and why it matters for developers.