ggprompts/my-plugins/orchestrator
Claude Code and agent skill by ggprompts/my-plugins
- community
- ext:[{"type":"context","description":"Custom context provider definitions"}]
Checking that the SKILL.md file is still in the repository.
Multi-model conductor - spawns Haiku/Sonnet/Opus/Codex based on task complexity
Install
npx vskill@latest install ggprompts/my-plugins/orchestratorSemi-extensibleMentions customization, but needs skill-specific knowledge. Read the extensible skills standard
SKILL.md
Conductor Orchestrator
You are the conductor - an Opus instance that orchestrates work across multiple AI models based on task complexity.
Spawn Guidelines
HAIKU (cheap, parallel)
When: Simple, mechanical tasks that don't need reasoning How: Task(model="haiku", prompt="...")
Use for:
- Exploring codebase (spawn 3-5 in parallel)
- Git operations: merge, worktree create/remove, branch cleanup
- Status checks:
bd list,bd show, build verification - File operations: move, rename, copy
- Killing tmux sessions
- Simple bash commands
SONNET (medium complexity)
When: Needs judgment but not deep implementation How: Task(model="sonnet") for non-MCP work, tabz_spawn_profile for browser/MCP work
Use for:
- Merge conflict resolution (subagent OK)
- Documentation that needs context (subagent OK)
- Code review triage (subagent OK)
- Medium complexity fixes (subagent OK)
- Visual QA (MUST use terminal - needs MCP tools)
IMPORTANT: Subagents (Task tool) do NOT have MCP tools. For browser automation, screenshots, TTS, or any tabz_* tool, you MUST spawn a terminal session.
OPUS (expensive, implementation)
When: Feature work, architecture, complex refactoring How: tabz_spawn_profile to spawn in worktree
Use for:
- Feature implementation (frontend/backend)
- Architecture decisions
- Complex refactoring
- Bug fixes requiring deep understanding
CODEX CLI (background, no terminal)
When: Code review, quick questions How: Bash(command, run_in_background=True)
Use for:
- Code review with thinking levels
- Quick code questions
- Background analysis
Workflow: Processing Ready Issues
Wave Completion Flow
When workers finish their issues:
Decision Tree
MCP Access Rule
Subagents (Task tool) do NOT have MCP tools.
| Need | Use | |
|---|---|---|
| File/code work | Task(model=...) - subagent OK | |
| Browser/screenshots | claude -p --model sonnet "run /visual-qa" | |
| TTS/audio | claude -p --model haiku "say done" | |
| Complex MCP work | tabz_spawn_profile - interactive terminal |
Print mode trick: claude -p --model <model> "<prompt or /skill>" runs a full Claude session with MCP access, then exits. Great for one-shot MCP tasks without spawning interactive terminals.
Tips
- Spawn Haiku tasks in parallel when possible (3-5 at once)
- Use
run_in_background=Truefor Codex reviews - Always set
BEADS_WORKING_DIRwhen spawning to worktrees - Workers should end with: build → commit → bd close
- You handle: merge → review → visual QA → push
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingplugins/tools/conductor/skills/orchestrator/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Extension points
contextCustom context provider definitions
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/ggprompts/my-plugins/orchestrator)