smart-agent-spawnerlisted

# Smart Agent Spawner ## Purpose Automatically select the optimal Claude model (Opus, Sonnet, or Haiku) and, where available, effort level, when spawning agents based on task complexity and agent type. This balances cost, speed, and capability, and defines what to do when a spawn fails. ## When This Skill Activates Applies automatically whenever you spawn agents via the `Agent` tool, `Task` tool, or `Workflow` tool's `agent()` calls. A `PreToolUse` hook (`hooks/model-router-nudge.sh`, matcher `Agent|Task|Workflow`) fires before every such call to remind you of these rules, so you don't need the user to ask. ## Model Selection Strategy ### Opus, most capable, highest cost Use for critical decisions that are hard to reverse or require deep reasoning. Agent types: - `Plan` - architecture and implementation planning - `code-reviewer` - code review and quality assessment - `code-security-auditor` - security reviews and vulnerability detection - `database-designer` - complex database schema design Task characteristics: - Architecture decisions and system design - Security reviews and threat modeling - Complex debugging (race conditions, memory leaks, subtle bugs) - Critical refactoring that affects many files - Code that handles sensitive data or authentication - Performance optimization requiring deep analysis - Migration planning between technologies Cost: the Opus tier runs roughly 1.7x Sonnet 5's $3/$15 per MTok, not the 5x gap older models had. Opus is also the defau