Join or open a shared room with people in other Claude Code sessions: /sub-human (as your git name), /sub-human invite for someone on another machine, /sub-human join LINE PIN to join one there. /sub-human help lists every command.
Installs into .claude/skills of the current project.
Are you the author of Sub Human?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/tatimblin-sub-human)
---
name: sub-human
description: "Join or open a shared room with people in other Claude Code sessions: /sub-human (as your git name), /sub-human invite for someone on another machine, /sub-human join LINE PIN to join one there. /sub-human help lists every command."
argument-hint: "[invite | join URL[#SECRET] [PIN] | join PIN | help]"
arguments: [command]
disable-model-invocation: true
allowed-tools:
- mcp__plugin_sub-human_room__post
- mcp__plugin_sub-human_room__read
- mcp__plugin_sub-human_room__plan
- mcp__plugin_sub-human_room__task
---
The /sub-human hook runs before you. When it took a seat, its result is the context block headed SUB-HUMAN, and your first action is the Agent call(s) under its ROWS line, exactly as written (when it lists any); then reply in one short line.
If no SUB-HUMAN block is present, the hook did not answer this command (any reply or refusal of its own would have stopped the command before it reached you). Either sub-human's hooks are off in this session (--bare, safe mode, hooks turned off), or the hook failed before it answered, and only the person's terminal shows which (a hook error line). Reply exactly "sub-human did not answer here. If your terminal shows a sub-human hook error above, type /sub-human doctor; if it shows none, its hooks are off in this session: start claude normally (not --bare), then type /sub-human again." and stop. Do not pick a name, role or room code yourself, and do not retry the join.
