Slash Commands

Type / in the chat input to open the slash-command picker. Claudette aggregates commands from four sources, deduplicates them, and ranks the picker by recency so what you used last is one keystroke away.

Built-in commands

Native commands ship with the app and always work. They split into two flavors:

Local actions and settings shortcuts — these don’t contact the agent; they mutate the UI or open a settings panel.

Command	Effect
`/clear`	Clear the chat history (UI-only).
`/compact`	Compact the conversation. Works on every harness. On Claude Code the CLI handles it natively; on Codex Native Claudette issues the `thread/compact/start` JSON-RPC; on the Pi SDK harness Claudette calls Pi’s native `AgentSession.compact()`. The timeline renders a compaction divider when the compaction completes. Pi also auto-compacts long sessions on its own — those land a divider too (labelled “automatic” or “context full”). A failed compaction surfaces a notice instead of a divider.
`/plan`	Toggle plan mode on for the next turn.
`/model` `[<model>]`	Show the active workspace model, or set it inline (`/model opus`, `/model claude-3-5-sonnet`, `/model pi/openai/gpt-5.4`, …).
`/permissions` `[readonly\|standard\|full]`	Show the current permission mode, or set it inline (`/permissions full`). Alias: `/allowed-tools`.
`/status`	Show session/permission status.
`/help`	Show available commands.
`/init`	Generate a `CLAUDE.md` for the current repo.
`/login`	Start sign-in for the active model provider. Claude models open the Claude Code browser/manual-code flow; Codex models run `codex login`; Pi models show `pi auth` guidance (run `pi auth` in a terminal, then refresh Pi models — Claudette does not query Pi auth status).
`/config` `[section]`	Open Claudette settings (optional section: `general`, `models`, `usage`, `appearance`, `notifications`, `git`, `plugins`, `experimental`).
`/usage`	Open the Claude Code usage panel.
`/extra-usage`	Open the extra-usage view.
`/release-notes`	Open the GitHub release notes.
`/version`	Show app + protocol version.

Prompt expansions — these seed pre-written prompts that flow through the agent:

Command	Effect
`/review [focus]`	Seed a code review of the current branch against its base.
`/security-review [focus]`	Seed a security-focused review of the current branch.
`/pr-comments [PR# or guidance]`	Summarize PR comments for the current branch.

Claude Code Plugins:

Command	Effect
`/plugin [install\|enable\|disable\|uninstall\|update\|manage\|browse\|marketplace]`	Open or drive Claude Code Plugins.
`/marketplace [add\|remove\|update] <source>`	Manage Claude Code marketplaces.

Custom commands

Drop a markdown file into one of these directories and it appears as a slash command:

Directory	Source	Scope
`<repo>/.claude/commands/`	`project`	Available only in this repository
`~/.claude/commands/`	`user`	Available everywhere

The file’s basename (without .md) becomes the command name. Frontmatter is optional:

---
description: Generate JSDoc for the selected file
argument-hint: <file path>
---

You are a JSDoc author. Generate complete JSDoc for the file passed as an argument:

@$ARGUMENTS

The file’s body is the seeded prompt. Use $ARGUMENTS to interpolate whatever the user typed after the command name.

Skills

A “skill” is a slash command backed by a directory rather than a single file — it can include supporting markdown, examples, and a SKILL.md that defines the command. Skills are discovered the same way commands are:

Directory	Source
`<plugin>/skills/`	Plugin-contributed
`~/.claude/skills/<skill-name>/SKILL.md`	User-contributed

The skill’s directory name is the command name; SKILL.md provides the seeded prompt and frontmatter.

When the agent pulls in a skill mid-turn, the chat transcript shows a ✦ <skill name> activated marker (wand icon) as its own entry — not bundled into the collapsible “N tool calls” group, the same way agent runs and thinking blocks stand on their own — so you can see which skills shaped a response at a glance.

Plugin-contributed commands

Enabled Claude Code plugins (managed via the Plugins settings panel) can ship their own commands and skills. They appear in the picker tagged with their plugin’s name. See the Claude Code documentation for plugin authoring; Claudette discovers them automatically once installed.

Resolution order

When the same command name is contributed by multiple sources, the highest-priority one wins:

Built-ins for /login, /plugin, and /marketplace — reserved; never shadowable.
Project commands (<repo>/.claude/commands/) — your repo’s customizations.
User commands (~/.claude/commands/).
Plugin commands.
Other built-ins (/review, /config, …) yield to a project- or user-defined command of the same name, so you can override them.

The picker shows the source label in muted text, so you can spot conflicts at a glance.

Recency ordering

The picker sorts by recency of last use, then alphabetical. Use Tab to autocomplete and Enter to dispatch.