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 organizing AI capabilities as ‘Skills’—folders with instructions, scripts, and knowledge—improves consistency, onboarding, and scalability. This approach shifts how organizations deploy and manage AI agents.

Anthropic has announced that its ‘Skills’ are not just prompts but structured folders containing instructions, scripts, and reference materials, a shift that enhances AI consistency and organizational knowledge management. This approach, based on internal experiments, aims to standardize how AI agents perform tasks across teams, making their operations more durable and scalable. The revelation highlights a fundamental rethinking of prompt engineering and AI asset management, with potential implications for enterprise AI deployment.

According to an internal write-up from a Claude Code engineer, Skills are now defined as folders that house instructions, reference documents, scripts, templates, data, configurations, and hooks. Instead of treating prompts as ephemeral text snippets, this method encapsulates the entire operational context in a reusable, versioned container. The agent can discover, read, and execute the scripts within these folders, enabling more reliable and consistent outputs.

Anthropic’s internal experiments involved running hundreds of Skills across their engineering teams, leading to a categorization into nine core types: library and API reference, product verification, data fetching and analysis, business process automation, code scaffolding, code review, CI/CD deployment, runbooks, and infrastructure operations. The company found that the most valuable Skills are those that verify outputs, reducing mistakes and improving quality.

By framing Skills as organizational assets rather than simple prompts, Anthropic emphasizes their role in onboarding, knowledge retention, and continuous improvement. The company allocates engineer time to refine each Skill, viewing them as assets that appreciate in value as they become more robust and comprehensive.

At a glance
reportWhen: published recently, with ongoing implem…
The developmentAnthropic published insights from running hundreds of Skills internally, emphasizing their nature as folders rather than prompts, and their impact on AI reliability and organizational processes.
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 Asset Management for Organizations

This development matters because it shifts the paradigm from ad-hoc prompting to structured, reusable organizational assets. By treating Skills as folders, companies can achieve greater consistency in AI outputs, accelerate onboarding of new team members, and create a scalable library of operational procedures. This approach addresses common challenges in enterprise AI deployment, such as variability, knowledge loss, and maintenance complexity. Ultimately, it could lead to more reliable, transparent, and manageable AI systems in business contexts.

AI Project Power: Reimagining Your Role in the Age of Artificial Intelligence

AI Project Power: Reimagining Your Role in the Age of Artificial Intelligence

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

From Prompt Engineering to Asset-Based AI Design

Prior to this, most organizations relied on prompt engineering—crafting specific instructions each time an AI task was performed. This approach is fragile, inconsistent, and hard to scale. Anthropic’s internal experiments with hundreds of Skills demonstrated that packaging knowledge into structured folders improves reliability and reusability. The concept aligns with broader trends in software engineering, where modular, version-controlled assets replace ephemeral scripts and notes. The nine categories identified by Anthropic serve as a blueprint for organizations to audit and enhance their AI capabilities, from basic reference management to complex operational workflows.

This shift is also a response to the limitations of prompt-based systems, which often require constant re-tuning and are prone to drift over time. By embedding institutional knowledge into Skills, organizations aim for AI systems that are more predictable and easier to maintain, especially as they scale.

“Treating Skills as assets—containers of knowledge—allows us to build more durable and consistent AI capabilities that can evolve over time.”

— Anthropic engineer

Unanswered Questions About Implementation and Scalability

It is not yet clear how widely Anthropic’s folder-based Skills approach has been adopted outside internal experiments or how it will perform at scale across diverse organizational contexts. Details about integration with existing workflows, tooling, and how organizations can transition from prompt-based to folder-based systems remain under development. Additionally, the long-term benefits and potential limitations of this approach are still being evaluated.

Next Steps for Broader Adoption and Validation

Anthropic plans to further develop and document its Skills framework, encouraging other organizations to adopt similar practices. Future updates may include tooling support for creating, managing, and versioning Skills as folders, as well as case studies demonstrating effectiveness at scale. Industry observers will watch for how this approach influences enterprise AI deployment and whether it becomes a standard best practice.

Key Questions

How does treating Skills as folders improve AI performance?

By encapsulating instructions, reference materials, and scripts in structured folders, Skills provide a more reliable and consistent operational context for AI agents, reducing variability and errors.

Can this approach be applied outside of Anthropic?

Yes, the concept is generalizable. Organizations can adopt folder-based Skills to standardize workflows, improve onboarding, and manage institutional knowledge more effectively.

What are the main categories of Skills identified?

The nine categories include library and API reference, product verification, data analysis, business automation, code scaffolding, code review, CI/CD, runbooks, and infrastructure operations.

What challenges might organizations face in adopting this method?

Potential challenges include developing tooling for managing Skills as folders, transitioning existing workflows, and ensuring proper version control and documentation practices.

Will this approach eliminate prompt engineering entirely?

Not necessarily; it aims to complement prompt engineering by providing a more structured, asset-based foundation for AI operations, making prompts more effective and consistent.

Source: ThorstenMeyerAI.com

This content is for general information only and is not financial, tax or legal advice. Consult a qualified professional for decisions about your money.
You May Also Like

The Delegation Ladder: The Four Agentic Loops, and What Each One Lets You Stop Doing

An analysis of the four agentic loops in AI development, detailing what each allows you to stop doing and how they influence AI process automation.

Kill-Switch-Proof: How to Build So Washington Can’t Take Your AI Stack Down

Learn the strategies to make AI stacks resilient against government shutdowns, including dependency mapping, gateways, fallback tiers, and open-weight models.

Saturation. The ten-essay framework, closed.

The ten-essay European AI landscape framework is now considered complete as of May 2026, with external developments expected to drive future insights.

The Ghost Story Became a Forecast.

Thorsten Meyer analyzes Jack Clark’s recent essay revealing a bivalent forecast for AI development, with 60% chance of rapid progress by 2028 and 40% indicating fundamental limits.