Registry Overview

The Kenji registry is a decentralized index of skills and stacks maintained across multiple public GitHub repositories. When you run kenji search or kenji install <namespace>/<name>, results are aggregated from the official registry and all connected community registries.

How registries work

Each registry is a GitHub repository containing only JSON metadata files. These files describe skills and stacks — they don't contain the actual Markdown instruction files. The repo field in each entry points to the repository where the actual content lives.

The list of all registered registries (official and community) is stored permanently in registry-map.json inside kenjiprotocol/registry on GitHub. This means the registry list is version-controlled and never lost due to cache expiry or redeployment.

The Kenji website and CLI search aggregate results from GitHub on a 30-minute cache cycle. When you first add a registry, the cache is immediately busted and rebuilt — your skills appear in search results right away. Subsequent edits to your registry files (descriptions, renames, new entries) are picked up automatically at the next refresh, within 30 minutes.

Official and community registries

The official Kenji registry is maintained at kenjiprotocol/registry — view it on GitHub ↗. Anyone can create and submit a community registry. Community skills are labeled in search results and require user confirmation before installing.

Search aggregation

Kenji queries its hosted API, which aggregates results across all registered registries. You don't need to specify which registry to search — all of them are included automatically.