Skip to content

amplihack RustyClawd — Command Reference

Synopsis

amplihack RustyClawd [OPTIONS] [ARGS]...

Description

Launches the RustyClawd tool via the native Rust launcher path. RustyClawd is a Claude-compatible tool, so it receives --dangerously-skip-permissions (when --skip-permissions is set) and --model injection from the launch flag injection system.

The command name is case-sensitive: use RustyClawd exactly (not rustyclawd or rusty-clawd). This is set via an explicit #[command(name = "RustyClawd")] attribute.

Options

Flag Type Default Description
--append <TEXT> string — Append instructions to a running auto mode session and exit. The session receives the text without restarting.
--no-reflection bool false Disable post-session reflection analysis.
--subprocess-safe bool false Skip shared launcher staging and environment updates for subprocess delegates. Use when RustyClawd is launched as a child of another amplihack process.
--auto bool false Run in autonomous agentic mode with iterative loop execution.
--max-turns <N> integer 10 Maximum number of turns for auto mode. Must be ≥ 1. Only meaningful when --auto is set.
--ui bool false Enable interactive UI mode for auto mode. Only meaningful when --auto is set.

Trailing Arguments

All positional arguments after the flags are forwarded verbatim to the RustyClawd/Claude binary:

amplihack RustyClawd --print 'Fix the failing tests'
# Forwards: --print 'Fix the failing tests'

Hyphen-prefixed values are allowed in trailing arguments.

Examples

# Launch RustyClawd interactively
amplihack RustyClawd

# Run in auto mode with 5 turns
amplihack RustyClawd --auto --max-turns 5

# Append to a running session
amplihack RustyClawd --append 'Also fix the linting errors'

# Launch without reflection analysis
amplihack RustyClawd --no-reflection

Flag Injection

RustyClawd is Claude-compatible. The launch flag injection system applies:

  • --dangerously-skip-permissions is injected when --skip-permissions is passed to the parent amplihack command
  • --model is injected unless the user supplies --model in trailing args

See Launch Flag Injection for details.

Exit Codes

Code Meaning
0 Success
1 Error