v3.2.0 — 10 Competitive Features

What's New in v3.2.0

10 New Features

Feature	Command	Description
Auto Lint-Fix	/lint auto	AI automatically fixes lint errors (max 3 iterations)
Auto Test-Fix	/test-fix <cmd>	Run tests → AI fixes → re-run loop (max 3 iterations)
Hooks System	config.json hooks	Pre/post hooks for any tool execution
Watch AI Triggers	# AI: instruction	Comments in watched files trigger AI actions
Architect Mode	/architect opus haiku	Strong model plans, cheap model implements
Headless Mode	--headless "task"	Run tasks non-interactively from CLI
Worktree Isolation	/agent --worktree	Git worktree per sub-agent
OS Sandboxing	/sandbox on	sandbox-exec (macOS), firejail/bwrap (Linux)
Repo Map	/map	Tree-sitter or regex symbol extraction
Agent Teams	/team <task>	Coordinator splits tasks, workers execute in parallel

Stats

• 69 slash commands (was 63)
• 284 tests passing (was 239)
• 11 Python modules
• ~8,500+ lines of code