Every name Claude Code reads.
Not the names we happened to write about. The names that are actually in the build, read out of the shipped bundle release by release. Eight kinds of them, 3,168 in all.
1,313 of them have no description from anybody: not a line in Anthropic's documentation, and not a changelog entry of ours that names them. They are in the build and nobody has ever said what they do.
The inventory below was last read out of v2.1.243. A name marked as gone was in an earlier build and is not in that one.
The eight families
claude command accepts, including the ones no help text mentions.
1,188 feature flags
Feature flags
GrowthBook gates the client evaluates on startup. Most of what ships is behind one of these before it is announced.
431 settings keys
Settings keys
Keys the client reads out of a settings.json, at whichever of the four scopes it finds one.
106 slash commands
Slash commands
Commands typed at the prompt, the built-in ones the bundle carries.
33 tools
Tools
The tools the model is given, by the name it calls them with.
31 hook events
Hook events
The points in a session a configured hook can be run at.
19 models
Models
Model identifiers the client knows about, including the aliases it resolves.
A family reading not read yet is one whose miner has not landed. The name is real, the route is real, and the list fills in on the release after the inventory learns to read it.