.well-known/agent-skills/index.json

What is the agent-skills catalog?

A JSON index in your .well-known folder listing the skills and capabilities agents should associate with your business, what you do, described in the vocabulary of the Agent Skills standard. At its core, an Agent Skill is a named, described capability; the catalog makes your business's capabilities discoverable at a predictable path.

Why AI agents need it

Skills are how the agent ecosystem now describes "things an agent can do or use." Agents that speak this vocabulary can read a skills catalog and immediately understand how to describe or engage a business, without parsing your whole site to guess at capabilities.

The business value for you

Your services become discoverable capabilities rather than paragraphs an agent may or may not interpret correctly. It is a small, static file that aligns your business with the fastest-moving capability vocabulary in the agent ecosystem.

How it came to be

Anthropic introduced Agent Skills in October 2025 and published the format as an open standard on December 18, 2025. The specification is community-maintained and has been adopted by a large number of AI tools and agentic clients. The core design idea, progressive disclosure, loads a skill's name and description first, and details only when needed, which is exactly the discovery pattern a static catalog serves.

Sources: Anthropic, Equipping agents for the real world with Agent Skills • Agent Skills specification (agentskills.io)

Adoption

The Agent Skills format itself has broad, fast adoption on the client side, the official spec repository gained over 22,000 GitHub stars within months of release, and the standard's client showcase lists a wide range of supporting tools. Publishing skills catalogs at a website's .well-known path is the newer, earlier-stage half: Cloudflare's 2026 agent-readiness analysis includes Agent Skills among the capability signals it checks, and found emerging capability standards present on only a small fraction of sites.

Sources: agentskills/agentskills repository • Cloudflare, Introducing the Agent Readiness score (2026)

The technical reality

Each skill entry needs a conformant name and description (the spec constrains both), and the catalog has to reflect capabilities your site actually supports. A skills index that overclaims is worse than none. Our generator derives skills from your analyzed site content, validates the structure, and keeps the catalog synchronized under the managed subscription. The upstream spec is young and evolving; tracking it quarterly is part of what you're paying us for.

What happens if you skip it

Agents fluent in the skills vocabulary find no catalog on your domain and fall back to inferring your capabilities from prose, the exact failure mode the standard was designed to eliminate.