Auto-mode classifier inspection commands.
Wraps claude auto-mode config|defaults|critique, which expose the
auto-mode classifier configuration the CLI uses to decide which actions
it can take without prompting.
Usage
config = ClaudeWrapper.Config.new()
# Print the effective config (your settings merged over defaults), as JSON
{:ok, json} = ClaudeWrapper.Commands.AutoMode.config(config)
# Print the default environment/allow/soft_deny/hard_deny rules, as JSON
{:ok, json} = ClaudeWrapper.Commands.AutoMode.defaults(config)
# Get AI feedback on your custom auto-mode rules (free-form text)
{:ok, feedback} = ClaudeWrapper.Commands.AutoMode.critique(config)
# Override the model used for the critique
{:ok, feedback} = ClaudeWrapper.Commands.AutoMode.critique(config, model: "opus")
Summary
Functions
Print the effective auto-mode config as JSON.
Get AI feedback on your custom auto-mode rules.
Print the default auto-mode environment, allow, soft_deny, and hard_deny rules as JSON.
Functions
@spec config(ClaudeWrapper.Config.t()) :: {:ok, String.t()} | {:error, term()}
Print the effective auto-mode config as JSON.
Emits your settings where set and defaults otherwise, merged. The string
is returned verbatim (trimmed); decode it with Jason.decode/1 if you
need the structured form.
@spec critique( ClaudeWrapper.Config.t(), keyword() ) :: {:ok, String.t()} | {:error, term()}
Get AI feedback on your custom auto-mode rules.
Output is free-form text.
Options
:model- Override which model is used for the critique (--model). Without one the CLI picks its default.
@spec defaults(ClaudeWrapper.Config.t()) :: {:ok, String.t()} | {:error, term()}
Print the default auto-mode environment, allow, soft_deny, and hard_deny rules as JSON.
Useful as a reference when writing custom rules. The soft/hard deny split distinguishes "warn but allow when explicitly overridden" from "always block."