Skip to content

Slash Commands Reference

All /z commands available inside Claude Code, listed alphabetically.

Command Description
/z adversary [profile] [--cycle] Switch the reviewer's behavioral profile (keeps context by default)
/z agent [agent_name] Switch Claude Code's active agent persona (cycles the reviewer)
/z ask Ask the reviewer a question without changing behavior
/z cycle Reset the reviewer's context window
/z direct Send a high-priority directive that overrides behavior
/z disable Disable supervision (current session or --all)
/z enable Enable supervision (current session or --all)
/z model [model_id] Switch the reviewer's model (always starts fresh)
/z monitor Open a live window streaming the reviewer's verdicts
/z plan <task> Hold a task under a planning gate until you agree an approach with the reviewer
/z restart Restart the Zwischen daemon
/z session-id Show the current session ID
/z spy View the reviewer's transcript for the current turn
/z status Show the reviewer's current model and profile
/z unblock Clear blocks without disabling supervision (current session or --all)
/z update Update the Zwischen client and restart the daemon

/z adversary [profile] [--cycle]

Switch the reviewer's behavioral profile. By default, the reviewer keeps its conversation context -- all settings update immediately (operating rules, tool visibility) without cycling the session. Pass --cycle to tear down the session and start fresh with a handoff summary.

If you provide a profile name directly (e.g. /z adversary code-review), the switch happens immediately without the wizard. Omit the profile name to see the full profile picker.

/z agent [agent_name]

Switch Claude Code's active agent persona mid-session. Presents a picker of agents from ~/.claude/agents/, resolves the reviewer profile mapped to the chosen agent in settings.yaml, and cycles the reviewer to a fresh session with the new agent's definition as its rubric. Claude Code then adopts the selected agent's instructions.

/z agent              # Agent picker wizard
/z agent accountant   # Switch immediately, no wizard

Unlike /z adversary (which changes only the reviewer's profile), /z agent changes both who Claude Code acts as and how the reviewer monitors it. The reviewer always cycles on an agent switch. The agent must have a matching profile in settings.yaml whose agents: list includes it.

/z ask

Ask the reviewer a question. Does not change monitoring behavior.

/z ask Why did you block that shell command?

/z cycle

Reset the reviewer's context window. The reviewer writes a handoff summary, then monitoring continues with a fresh context.

/z direct

Send a directive that overrides the reviewer's behavior for the rest of the session.

/z direct Let the agent write to config files without review.

/z disable / /z enable

Toggle supervision on or off.

/z disable          # This session
/z disable --all    # All sessions
/z enable           # Re-enable this session
/z enable --all     # Re-enable all sessions

/z model [model_id]

Switch the reviewer's model mid-session. This always cycles the session – the reviewer writes a handoff summary, then a new session starts on the target model with that summary plus a tail of recent events. The same cycle settings (tail_depth, etc.) that govern automatic and manual cycling apply here.

/z model gemini-3.1-pro-preview    # Switch immediately
/z model                           # Model picker wizard

When called without arguments, the skill checks which providers have usable credentials (via zwischen providers), shows a filtered model list grouped by provider, and prompts for selection. Providers without configured credentials are omitted. For OpenAI, you're prompted to choose between OAuth and API key if both are available.

Credentials for the target model must be available. If they aren't, the command tells you to run zwischen auth.

/z monitor

Open a separate terminal window that streams the reviewer's verdicts live, alongside your session. Live-only – it shows verdicts from the moment it opens, not history. See Live Monitor.

/z plan <task>

Hold a task under a planning gate: the agent must agree a rough approach with the reviewer before it starts real work. Everything after plan is the task.

/z plan Refactor the auth middleware to support API keys

This places the hold regardless of the profile's planning_gate setting. If the agent is already blocked, clear it with /z unblock first.

/z restart

Restart the Zwischen daemon without updating.

/z session-id

Print the current session ID. Useful for diagnostics and support.

/z spy

View the reviewer's transcript. Shows reviewed events, determinations, and reasoning.

/z spy          # Latest turn
/z spy 3        # Turn 3

/z status

Show the reviewer's current model and behavioral profile for this session.

gpt-5.5 (openai) · general

/z unblock

Clear all active blocks for the current session without disabling supervision. The reviewer keeps watching; the current block is just released.

/z unblock          # This session
/z unblock --all    # All sessions

Use this when message_adversary is unavailable and you need to clear a block without turning off monitoring entirely. Unlike /z disable, supervision continues after unblocking.

/z update

Update the zwischen package to the latest version and restart the daemon.