Create Brain

Skill: /create-brain

Model preference: #lowcost (per horizon_aios_model_prefs.md; overridable by a prompt directive).

Provision a new isolated AI brain with horizon_aios_create_brain.py: an OS user account, the shared brains group, a per-brain group, the workspace at $HORIZON_ROOT/brains/<brain-name>/, a login shell profile, and an auto-generated 64-char password stored in the OS native keystore. The sbin/skills_sbin/logs Deny ACEs are re-applied after all grants per the security invariants.

When to invoke

The user types /create-brain, or asks to "create a brain", "provision a brain account", "stand up a new brain", or "add a brain user".

Arguments

/create-brain <brain-name> [--automation none|scheduled|daemon] [--horizon-root PATH] [--dry-run]

• <brain-name> — required positional; must match ^[a-z][a-z0-9_]{1,19}$ (start with a lowercase letter, then 1–19 lowercase letters/digits/underscores; max 20 chars — Windows local user name limit). If the user did not supply one, ask before running. If the supplied name exceeds 20 characters, reject it and ask for a shorter name before running.
• --automation — opt-in logon-rights tier (default none). scheduled grants "Log on as a batch job" (Windows) / enables systemd lingering (Linux); daemon grants "Log on as a service" (Windows) / prints system-service guidance.
• --horizon-root PATH — explicit root; otherwise derived from the script location.
• --dry-run — print every action without making changes.

Step-by-step execution

Step 1 — Check elevation

This hard-requires the administrative context — the script exits with an error if not run as Administrator (Windows) or root (Unix). If the session is not elevated and the user did NOT pass --dry-run, do not run it: tell the user to re-run from an elevated terminal (or with sudo) and stop. --dry-run still calls the same privilege check, so a real preview also needs elevation; if blocked, surface the script's own message.

Step 2 — Run

python "$HORIZON_SYSTEM/sbin/horizon_aios_create_brain.py" <brain-name>

Append --automation / --horizon-root / --dry-run as requested.

Step 3 — Report

3.1 Relay the Phase 4 verification checks ([PASS]/[FAIL]) and the Summary.

3.2 Surface the "Next steps" the script prints — notably retrieving the account password: python "$HORIZON_SYSTEM/sbin/horizon_aios_brain_credential.py" get <brain-name> --show. The password is never printed by provisioning; it lives in the OS keystore.

3.3 If verification reported failures, relay the script's cleanup instructions (it points at horizon_aios_remove_brain.py <name> --yes).

Notes for the executing agent

• The generated password is stored in the OS native keystore via horizon_aios_brain_credential.py and is NEVER echoed. Do not attempt to print or log it.
• If the user already exists, the script reports "Nothing to do" and exits 0 — that is not a failure.
• Per-brain group naming: <brain-name>_group on Windows (shared user/group namespace), <brain-name> on Linux/macOS.
• Deny ACEs on sbin/skills_sbin/logs are always (re)applied AFTER all brains grants — do not reorder; just run the script.