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 for AI agents should be conceived as folders containing instructions, scripts, and assets, not just prompts. This approach enhances consistency, onboarding, and scalability in AI workflows. The company ran hundreds of Skills internally to refine this methodology.

Anthropic has revealed that its internal approach to building AI agent capabilities involves treating Skills as folders—containing instructions, scripts, data, and configurations—rather than just saved prompts. This shift aims to create durable, reusable organizational assets that improve consistency and onboarding, marking a significant evolution in enterprise AI deployment.

In a detailed write-up from a Claude Code engineer, Anthropic explained that Skills are fundamentally folders—not prompts—containing a variety of assets that enable AI agents to perform complex tasks reliably across the organization. This redefinition allows agents to discover, read, and execute scripts and instructions stored within these folders, making their behavior more predictable and easier to update.

Anthropic’s internal experiments involved running hundreds of Skills across different engineering teams, leading to the development of a nine-category taxonomy. These include reference libraries, verification procedures, data analysis, automation workflows, code scaffolding, review processes, deployment routines, runbooks, and infrastructure operations. The company found that focusing on verification Skills, which check the work done by other Skills, delivered the most significant quality improvements.

This approach emphasizes that Skills are not just static prompts but assets that evolve over time, becoming more refined with each iteration. Anthropic advocates dedicating engineer time to perfect a single Skill category, viewing Skills as an appreciating organizational resource rather than a cost.

At a glance
reportWhen: published recently, based on internal A…
The developmentAnthropic published insights from its internal use of Skills, showing they are structured as folders with instructions and assets, not simple prompts, to improve AI agent performance and organization-wide consistency.
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 Workflows with Folder-Based Skills

This development matters because it shifts the paradigm from ad-hoc prompting to a structured, asset-based approach that enhances reliability, scalability, and knowledge retention within organizations. By encapsulating tribal knowledge and guardrails in Skills folders, companies can standardize AI behavior, reduce onboarding time, and create a library of refined, reusable components. This approach also encourages continuous improvement, as Skills evolve through iterative refinement, turning organizational knowledge into a durable asset.

Amazon

AI workflow management software

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

From Prompting to Asset Management in AI Development

Prior to this insight, most teams relied on manually retyping prompts daily, which led to inconsistent outputs and poor knowledge transfer. Anthropic’s internal experiments, detailed in their recent publication, demonstrate a shift towards treating Skills as comprehensive containers for operational procedures, scripts, and documentation. This approach aligns with broader trends in enterprise AI, emphasizing automation, reliability, and knowledge management. The concept builds on existing practices but elevates them into a formalized asset management system, similar to version-controlled code repositories.

Anthropic’s nine-category Skills map provides a framework for organizations to identify gaps in their AI capabilities and prioritize development efforts—ranging from reference libraries to operational runbooks—aiming to create a comprehensive, maintainable AI toolkit.

“Reframing Skills as folders radically changes how organizations design and deploy AI agents, transforming them into durable assets rather than fleeting prompts.”

— Thorsten Meyer, AI researcher

Amazon

AI agent scripting tools

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Unclear Aspects of Folder-Based Skills Adoption

It is not yet clear how widely other organizations will adopt this approach or how it will scale beyond Anthropic’s internal use. The long-term impact on AI reliability and maintenance costs remains to be seen, as does the integration with existing enterprise systems. Additionally, the precise technical implementation details, such as scripting standards and version control practices, are still evolving and have not been fully disclosed.

Amazon

enterprise AI automation platform

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Next Steps for Broader Adoption and Refinement

Organizations interested in this approach should evaluate their current AI workflows and consider developing their own Skills libraries, focusing on verification and automation. Anthropic plans to publish more detailed technical guidance and case studies to facilitate adoption. Future developments may include standardized frameworks for Skills management, integration with enterprise DevOps pipelines, and further refinement of the nine-category taxonomy.

Amazon

AI development environment

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Key Questions

What exactly is a Skill in Anthropic’s framework?

A Skill is a folder containing instructions, scripts, reference materials, and configurations that enable an AI agent to perform specific tasks reliably within an organization.

How does this differ from traditional prompting?

Unlike prompts, which are just text instructions, Skills are structured assets—folders—that contain comprehensive resources and scripts, making them reusable and easier to update over time.

Why is this approach considered more durable?

Because Skills encapsulate tribal knowledge and operational procedures as assets, they can evolve, be versioned, and reused, reducing reliance on ad-hoc prompts and enabling consistent performance.

What are the main categories of Skills identified by Anthropic?

They include reference libraries, verification routines, data analysis, automation workflows, code scaffolding, review processes, deployment routines, runbooks, and infrastructure operations.

What are the potential challenges of adopting folder-based Skills?

Challenges may include technical complexity in implementation, integrating with existing systems, establishing standards for scripting and versioning, and ensuring organizational buy-in for the new paradigm.

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

Rebrandable client delivery dashboard for AI agencies

A new rebrandable client delivery dashboard for AI agencies is being tested to improve client transparency and professionalism, with early validation underway.

The Compounding Error Problem — Why 99.9% Alignment Decays to 60% in 500 Generations

Research shows that even 99.9% accuracy in AI alignment can decline sharply over multiple generations, risking control loss in recursive self-improvement.

China: The Visible Hand

China’s government actively directs AI, robotics, and industrial growth through top-down planning, emphasizing state ownership and control, with private firms playing a supporting role.

ShinyHunters · The New APT Model.

Analysis of ShinyHunters’ evolving operational model, AI-enabled tactics, and implications for enterprise cybersecurity post-2026.