Methodology

How an entry gets listed

We find candidates in open sources: public skill catalogs, the official MCP registry, plugin marketplaces and GitHub topics. From other catalogs we take only the link to the original source. Descriptions are written by us from the author’s README and SKILL.md.

Anyone can submit a skill through the submission form. No account is needed.

• The entry has a reachable source: a repository, a package or a server URL.
• It is clear what the tool does and which agent it is for.
• The author is identifiable and the repository is not empty or silently abandoned.
• Listing in the catalog is never sold.

Where the data comes from

• Stars, license and last commit date come from the GitHub API and update automatically. Stars belong to the whole repository, so monorepos share one number.
• The copy counter counts clicks on the copy button on the site. We cannot see installs inside agents and do not show invented numbers.
• Install commands come from the author’s documentation. If the author gives no command, we say so.

Risk level

The risk level describes what a skill or server can do on your computer or in your accounts. It is not a quality score.

• Low: text instructions for the agent only, no programs or network calls.
• Medium: needs API keys, makes network calls or runs scripts within clear limits.
• High: arbitrary shell commands, downloaded binaries, broad access to accounts, money or data.

Editors’ pick

The editors’ pick badge means the project is widely used, well documented and maintained. It is an editorial opinion, not a measurement.

Translations

The primary catalog language is Russian. The English version is machine translated with a domain glossary and spot checked by the editors. If a translation is not ready, the English page shows the Russian text with a note.

Incidents

If an entry turns out to have a serious vulnerability or malicious code, we unpublish it and explain why. Use the report button on any entry page to tell us about a problem.