Kudosflow2 v1.2.0 - Agent Teams: Dynamic Multi-Agent Orchestration

What's New in v1.2.0

Added

• README: Added "Testing Agent Teams" section with verification criteria and 7 test cases (T-01 through T-07) covering marketing research, academic survey, travel planning, content creation, data analysis, legal/compliance, and code generation domains

Changed

• Packaging: Added scripts folder to the extension package
• Teams leader.json systemPrompt: Replaced SKILL.md delegation with fully embedded step-by-step orchestration instructions (Step 1–8) including worker kill, template load, JSON write, launch, healthcheck, parallel task send, and result integration
• Teams leader.json leader_node: Added injectSkillsPrompt: false to prevent skills prompt injection; added debug logging for message count, types, tool_calls, and response content preview
• Teams leader.json tools_node: Added "teams" to excludeTools to prevent recursive self-invocation
• Teams finalize_node: Included report content in interrupt message so results can be reviewed before the yes/no prompt
• Teams finalize_node: Changed final assistant message to show only cleanup result, eliminating duplicate report display
• Teams / SKILL.md: Translated all Japanese text in JSON configs, skill files, and scripts to English
• Rename: Renamed swarms → teams (folders, files, and all references) to align with Claude Code official terminology
• Teams leader systemPrompt: Embedded full step-by-step orchestration instructions directly in the system prompt for more reliable execution
• Removed: skills/teams/SKILL.md — no longer needed as instructions are now embedded in the system prompt
• task-creation.json: Translated Japanese regex patterns and string literals to English