A Skill Is a Folder, Not a Prompt: What Anthropic Learned Running Hundreds of Them

📊 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 Skills are best understood as folders containing instructions and assets, not just prompts. This approach improves consistency, onboarding, and scalability of AI agents, marking a shift in how organizations deploy AI capabilities.

Anthropic has publicly detailed its approach to building AI agent Skills as reusable folders, marking a significant shift from the common prompt-based methods. This development, based on internal experiments, emphasizes that Skills are not just saved prompts but comprehensive containers that include instructions, scripts, reference documents, and configuration. This approach aims to standardize and scale AI deployment within organizations, making agent behavior more consistent and maintainable.

According to a write-up from Anthropic’s Claude Code team, a Skill is a folder that contains a variety of assets—instructions, code, data, and hooks—that an AI agent can discover, read, and execute. This reframing moves away from viewing Skills as simple text prompts, instead positioning them as structured containers that encapsulate organizational knowledge and operational procedures.

Anthropic’s internal experiments involved running hundreds of Skills across its engineering teams, leading to a taxonomy of nine key categories, including verification, data fetching, automation, and infrastructure. The company found that well-designed Skills improve output consistency, reduce onboarding time, and allow continuous improvement through iteration.

Anthropic emphasizes that investing engineering effort into refining Skills, especially in verification, yields high returns by catching mistakes and improving output quality. The approach also facilitates versioning, sharing, and maintaining institutional knowledge, transforming ad-hoc prompting into durable, scalable capabilities.

At a glance
reportWhen: announced March 2024
The developmentAnthropic published new insights from its internal use of Skills, revealing that Skills are containerized folders with instructions, scripts, and assets, not simple prompts.
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

Transforming AI Deployment with Containerized Skills

This development matters because it shifts the paradigm from ephemeral prompts to durable organizational assets. By treating Skills as folders with embedded instructions, scripts, and knowledge, companies can achieve greater consistency, reduce onboarding time, and build a scalable library of operational capabilities. This approach could lead to more reliable and maintainable AI systems, especially as organizations scale their AI use across complex workflows.
Your Company Mandated AI: Instructions Not Provided

Your Company Mandated AI: Instructions Not Provided

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

From Prompt Engineering to Organized Asset Management

Traditionally, organizations have relied on prompt engineering—crafting specific instructions for AI models each time they are used. This method can be inconsistent and hard to scale. Anthropic’s new approach, inspired by internal experiments, treats Skills as structured containers that encapsulate organizational knowledge, operational procedures, and code. This shift reflects a broader move toward integrating AI more deeply into business processes, with a focus on standardization and asset management. The concept of Skills as folders emerged from Anthropic’s internal efforts to improve agent reliability and efficiency, with their best Skills evolving through iterative refinement and documentation.

“This reframing of Skills as folders containing instructions and assets fundamentally changes how organizations can design and maintain AI capabilities.”

— Thorsten Meyer, AI researcher

Agent-Based Software Development (Agent-Oriented Systems)

Agent-Based Software Development (Agent-Oriented Systems)

Used Book in Good Condition

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

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 perform in different organizational contexts. Details about the technical implementation, such as how agents discover and prioritize Skills in complex environments, remain under development. Additionally, the long-term impact on AI reliability and maintenance is still being evaluated.
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 Scaling and Validating the Skills Approach

Anthropic plans to further develop its Skills library, refine categorization, and test scalability across different teams and use cases. Other organizations may begin experimenting with containerized Skills, and industry standards could emerge. Monitoring how this approach influences AI reliability and operational efficiency will be key in the coming months.
Amazon

AI automation and configuration folders

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Key Questions

How is a Skill different from a prompt?

A Skill is a folder containing instructions, scripts, and assets that define how an AI agent performs a task, whereas a prompt is a simple instruction or question sent to the model. Skills are reusable, versioned containers that encode organizational knowledge, not just a one-time prompt.

Why does treating Skills as folders matter for organizations?

This approach allows organizations to standardize behaviors, improve consistency, reduce onboarding time, and continuously improve capabilities through iterative refinement of the Skills library.

Will this approach work with all types of AI tasks?

While initially tested in coding and operational workflows, the containerized Skills model has potential for broader application, but its effectiveness in diverse domains remains to be validated.

What are the technical challenges of implementing Skills as folders?

Challenges include designing discovery mechanisms, managing version control, and ensuring that the agent correctly interprets and executes the assets within each Skill, especially in complex environments.

Could this approach replace prompt engineering entirely?

It is unlikely to replace prompt engineering entirely but offers a more scalable, maintainable alternative for organizations seeking reliable, repeatable AI behaviors.

Source: ThorstenMeyerAI.com

You May Also Like

A War Room for Your Next Idea: Inside IdeaClyst

Discover how IdeaClyst provides founders with a local-first AI war room to validate ideas, reduce costs, and make better strategic decisions in real-time.

When One Agent Isn’t Enough: Claude Now Builds Its Own Team Of Agents On The Fly

Anthropic’s Claude now autonomously assembles its own agent teams for complex tasks, enhancing performance on high-value projects.

The Ghost Story Became a Forecast.

Thorsten Meyer analyzes Jack Clark’s recent essay, revealing a bivalent forecast: 60% chance of automated AI R&D by 2028, 40% indicating fundamental paradigm limits.

The Humanoid Robotics Reality Check: Q2 2026 Pilot-to-Production Status

Humanoid robots are shipping at scale in China, but Western deployments remain largely pilot-stage, with production ramping in 2026. The industry shows progress but also structural divides.