diverter-modelisted

# Diverter Mode Manage Diverter's user-level Delegation Policy. This is an explicit control surface, not a delegation task. Never invoke `$diverter`, evaluate a lineup, or spawn subagents while processing this skill. ## Accepted Operations - `auto` — persist automatic dispatch for suitable tasks - `ask` — persist approval-first delegation - `status` — report the policy that the next `SessionStart` will load - no argument — behave like `status` and also list the three operations above Reject every other value without changing configuration. Do not guess aliases or accept configuration paths. ## Run Mode Control Let `skill_file` be the fully expanded absolute path to this `SKILL.md`, compute `plugin_root = Path(skill_file).parents[2]`, and verify `${plugin_root}/.codex-plugin/plugin.json` exists. Run exactly one of: ```bash python3 "${plugin_root}/scripts/diverter-mode.py" auto python3 "${plugin_root}/scripts/diverter-mode.py" ask python3 "${plugin_root}/scripts/diverter-mode.py" status ``` Parse the JSON on stdout and require a zero exit status before reporting success. ## User Response After a successful `auto` or `ask` operation, say exactly: ```text Diverter mode changed to `<policy>`. Restart or reopen the task to apply the new mode. ``` For `status`, report the returned policy as the configuration that the next `SessionStart` will load. If `source` is `default`, say that `ask` is the default because configuration is missing. If `warning` is `invalid_config`,