- Dispatch timing: Commands are processed by the CLI before your prompt is sent to the model, allowing them to modify session state, display information, or trigger actions without consuming context.
- Aliases: Many commands have shorter aliases (e.g.,
/quit → /exit, /new → /clear, /rc → /remote-control). Both forms work identically.
- Environment variable disablers: Some commands can be disabled via
DISABLE_*_COMMAND env vars (e.g., DISABLE_DOCTOR_COMMAND=1 hides /doctor). Check each command’s entry for its disabler.
- Provider gating: Commands like
/setup-bedrock and /setup-vertex require the corresponding provider env var set (CLAUDE_CODE_USE_BEDROCK=1, CLAUDE_CODE_USE_VERTEX=1) to appear.
- Plan and tier restrictions: Some commands are Pro/Max only (
/privacy-settings, /passes, /upgrade). Others are free tier only. Subscription status is checked at runtime.
- Experimental feature flags: Commands like
/team-onboarding, /tasks (with full persistence), and Agent Teams require CLAUDE_CODE_* feature flags or internal Statsig flags to activate.
- [Skill] commands: These are bundled prompt-based skills (e.g.,
/batch, /loop, /debug, /simplify, /claude-api, /less-permission-prompts) that behave like slash commands but are implemented in markdown, not in the CLI binary. They can be disabled, customized, or extended like any skill.
← Back to Commands/README.md
