snow-flow 4.3.4 → 4.3.6

package/dist/cli.js

@@ -816,10 +816,13 @@ TodoWrite([
 ### 4. Agent Spawning Strategy
 Based on the task analysis, spawn ${taskAnalysis.estimatedAgentCount} agents in smart batches:
 
-**Agent Spawn Order:**
-1. **Primary Agent**: Spawn ${taskAnalysis.primaryAgent} first
-2. **Supporting Agents**: Spawn ${taskAnalysis.supportingAgents.join(', ')} after primary is established
-3. **Use Task tool**: \`Task("agent description", "agent prompt")\` for each agent
+**Dynamic Agent Discovery & Spawning (Use Snow-Flow MCP Tools):**
+1. **Initialize Swarm**: \`swarm_init({ topology: 'hierarchical', maxAgents: ${parseInt(options.maxAgents)} })\`
+2. **Discover Agents Dynamically**: \`agent_discover({ task_analysis: ${JSON.stringify(taskAnalysis)}, required_capabilities: [], context: { max_agents: ${parseInt(options.maxAgents)}, include_new_types: true } })\`
+3. **Spawn Discovered Agents**: Use \`agent_spawn({ type: 'agent_type_from_discovery', name: 'Agent Name', capabilities: ['discovered_capabilities'] })\` for each discovered agent
+4. **Coordination**: \`task_orchestrate({ task: '${objective}', strategy: 'adaptive' })\`
+
+**IMPORTANT**: Do NOT use hardcoded agent types like 'widget-creator', 'specialist' etc. Let \`agent_discover\` determine the optimal agent types for this specific task!
 
 ### 5. Memory Coordination Pattern
 All agents MUST use this simple memory coordination:

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "snow-flow",
-  "version": "4.3.4",
+  "version": "4.3.6",
   "description": "Conversational ServiceNow development platform using Claude Code. Multi-agent orchestration with 20+ MCP servers providing 200+ ServiceNow tools for comprehensive platform development.",
   "main": "dist/index.js",
   "type": "commonjs",