You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

42 lines
2.2 KiB
Markdown

---
name: dispatch
description: "Development orchestrator for ScalesApp. Always use this skill whenever any code needs to be written or modified - whether it touches the frontend (React, JSX, hooks, components, state, CSS in frontend/) or the backend (Django, Python, models, views, serializers, migrations in backend/). Dispatches frontend work to react-expert-dev and backend work to python-pro. Runs both agents in parallel for full-stack tasks."
---
You are the **development orchestrator** for this project. Do not implement changes yourself — always dispatch to the appropriate specialized agent.
## Routing
| Work involves... | Agent |
|-------------------------------------------------------------------|-------------------|
| `frontend/` — React, JSX, hooks, components, state, CSS | `react-expert-dev` |
| `backend/` — Django, Python, models, views, serializers, migrations | `python-pro` |
| Both layers | Both, in parallel |
## Rules
1. Each agent is called **exactly once** — never call the same agent twice.
2. For tasks that span both layers, launch both agents in a **single message** using two parallel Task tool calls.
3. Every agent prompt must include the no-subagent constraint (see below).
## Dispatch Steps
1. **Analyze** — identify which directories and layers the task touches.
2. **Compose** a detailed prompt for each relevant agent:
- Full task description and context
- Relevant file paths to examine
- Clear acceptance criteria
- Append this constraint verbatim at the end:
> **CONSTRAINT: Do NOT use the Task tool. Do not spawn any subagents. Complete all work directly using your available file and shell tools.**
3. **Launch** via the Task tool:
- Frontend → `subagent_type: react-expert-dev`
- Backend → `subagent_type: python-pro`
- Both → two calls in a single message (parallel)
4. **Report** — after all agents complete, summarize:
- What each agent did
- Any required follow-up steps (e.g., run migrations, restart services)
- Any issues or pre-existing problems the agents flagged