Orchestrates Stage 0 research and planning for JUCE plugins - creates architecture.md and plan.md contracts through subagent delegation. Use when creative brief exists and plugin needs DSP architecture specification, complexity assessment, or implementation planning. Invoke via /plan command, natural language (plan architecture, research DSP), or as first implementation step.
Content & Writing
100 Stars
38 Forks
Updated Nov 19, 2025, 01:13 PM
Why Use This
This skill provides specialized capabilities for glittercowboy's codebase.
Use Cases
Developing new features in the glittercowboy repository
Refactoring existing code to follow glittercowboy standards
Understanding and working with glittercowboy's codebase structure
---
name: plugin-planning
description: Orchestrates Stage 0 research and planning for JUCE plugins - creates architecture.md and plan.md contracts through subagent delegation. Use when creative brief exists and plugin needs DSP architecture specification, complexity assessment, or implementation planning. Invoke via /plan command, natural language (plan architecture, research DSP), or as first implementation step.
allowed-tools:
- Read # For contracts and references
- Write # For architecture.md, plan.md
- Edit # For state updates (PLUGINS.md, .continue-here.md)
- Bash # For git commits, precondition checks
- Task # For subagent delegation
- WebSearch # For professional plugin research
- Grep # For searching existing implementations
- Glob # For finding reference files
preconditions:
- creative-brief.md must exist in plugins/[Name]/.ideas/
- Parameter specification must exist (parameter-spec.md OR parameter-spec-draft.md)
- Plugin must NOT already be past Stage 0
---
# plugin-planning Skill
**Purpose:** Handle Stage 0 (Research & Planning - consolidated) through subagent delegation. This skill creates the foundation contracts (architecture.md, plan.md) that guide implementation in a single consolidated pass.
**Invoked by:** `/plan` command or as first step of `/implement` workflow
---
## Entry Point
**Check preconditions first:** See [references/preconditions.md](references/preconditions.md) for detailed validation logic.
Quick validation:
1. creative-brief.md must exist at `plugins/[Name]/.ideas/`
2. Parameter specification required (parameter-spec.md OR parameter-spec-draft.md)
3. Plugin status must be ≤ Stage 0 (not already in implementation)
4. Detect existing contracts (architecture.md, plan.md) for resume logic
If all preconditions pass → proceed to Stage 0
---
## Stage 0: Research & Planning
**Goal:** Create DSP architecture specification (architecture.md) AND implementation plan (plan.md)
**Duration:** 5-35 minutes (complexity-dependent)
**Implementation:** Delegated to research-planning-agent subagent via Task tool
**Dispatch pattern:**
1. Read contracts in parallel (use parallel Read calls in single tool invocation):
- plugins/[Name]/.ideas/creative-brief.md
- plugins/[Name]/.ideas/parameter-spec.md OR parameter-spec-draft.md
- plugins/[Name]/.ideas/mockups/*.yaml (if exists)
2. Invoke subagent:
```
Task(subagent_type="research-planning-agent", description="[prompt with contracts]", model="sonnet")
```
See [references/subagent-invocation.md](references/subagent-invocation.md) for detailed prompt construction.
3. After subagent returns, execute checkpoint protocol:
- Read subagent's return message
- Verify architecture.md and plan.md created
- Present decision menu (use assets/decision-menu-stage-0.md template)
- WAIT for user response
**Subagent executes:** Full Stage 0 protocol from research-planning-agent.md
**Outputs:**
- `plugins/[Name]/.ideas/architecture.md` (see assets/architecture-template.md)
- `plugins/[Name]/.ideas/plan.md` (see assets/plan-template.md)
**Decision menu:** Use assets/decision-menu-stage-0.md template
**Menu option routing:**
- **Option 1 (Start implementation):** Invoke plugin-workflow skill
- **Option 2 (Review contracts):** Read and display architecture.md, plan.md, parameter-spec.md, creative-brief.md
- **Option 3 (Improve brief):** Invoke plugin-ideation skill
- **Option 4 (Deep research):** Invoke deep-research skill
- **Option 5 (Pause):** Verify state saved, exit skill
- **Option 6 (Other):** Route based on request (plugin-workflow, ui-mockup, deep-research, or handle directly)
**Note:** research-planning-agent runs in fresh context (5-35 min session doesn't pollute orchestrator)
**Validation before proceeding to Stage 1:**
1. Verify architecture.md exists with all required sections (Core Components, Processing Chain, Parameter Mapping)
2. Verify plan.md exists with complexity score and implementation strategy
3. If verification fails: Display error and return to Stage 0
4. When user chooses option 1: Delegate to plugin-workflow skill (plugin-workflow owns Stages 1-4)
---
## Reference Files
- [references/preconditions.md](references/preconditions.md) - Detailed precondition validation logic
- [references/subagent-invocation.md](references/subagent-invocation.md) - Subagent dispatch pattern and prompt construction
- [references/state-updates.md](references/state-updates.md) - State file update protocol
- [references/git-operations.md](references/git-operations.md) - Commit patterns
**Templates:**
- `assets/architecture-template.md` - DSP architecture contract template
- `assets/plan-template.md` - Implementation plan template
- `assets/decision-menu-stage-0.md` - Decision menu template