prd

Prompt file imported from JacobCoffee/debug-toolbar (.claude/commands/prd.md). Fill in {{arguments}} before use. Copyright stays with the author.

Intelligent PRD Creation Workflow

You are creating a Product Requirements Document for: {{arguments}}

Intelligence Layer (ACTIVATE FIRST)

Before starting checkpoints:

1. Read MCP Strategy: Load .claude/mcp-strategy.md for tool selection
2. Learn from Codebase: Read 3-5 similar implementations
3. Assess Complexity: Determine simple/medium/complex
4. Adapt Workflow: Adjust checkpoint depth

Critical Rules

1. CONTEXT FIRST - Read existing patterns before planning
2. NO CODE MODIFICATION - Planning only
3. PATTERN LEARNING - Identify 3-5 similar features
4. ADAPTIVE DEPTH - Simple=6, Medium=8, Complex=10+ checkpoints
5. RESEARCH GROUNDED - Minimum 2000+ words research
6. COMPREHENSIVE PRD - Minimum 3200+ words

Checkpoint 0: Intelligence Bootstrap

Load project intelligence:

1. Read CLAUDE.md
2. Read specs/guides/patterns/README.md
3. Read .claude/mcp-strategy.md

Learn from existing implementations:

For panels: Read files in src/debug_toolbar/core/panels/ For integrations: Read files in src/debug_toolbar/litestar/ For extras: Read files in src/debug_toolbar/extras/

Assess complexity:

• Simple: Config change, bug fix, single file → 6 checkpoints
• Medium: New panel, API endpoint, 2-3 files → 8 checkpoints
• Complex: New integration, architecture change, 5+ files → 10+ checkpoints

Output: "✓ Checkpoint 0 complete - Complexity: [level], Checkpoints: [count]"

Checkpoint 1: Pattern Recognition

Identify similar implementations:

1. Search for related patterns in src/debug_toolbar/
2. Read at least 3 similar files
3. Extract naming patterns
4. Note testing patterns from tests/

Document:

## Similar Implementations

1. `src/path/to/similar1.py` - Description
2. `src/path/to/similar2.py` - Description
3. `src/path/to/similar3.py` - Description

## Patterns Observed

- Class structure: ...
- Naming conventions: ...
- Error handling: ...

Output: "✓ Checkpoint 1 complete - Patterns identified"

Checkpoint 2: Workspace Creation

mkdir -p specs/active/{slug}/research
mkdir -p specs/active/{slug}/tmp
mkdir -p specs/active/{slug}/patterns

Create initial files:

• specs/active/{slug}/research/plan.md
• specs/active/{slug}/tmp/new-patterns.md

Output: "✓ Checkpoint 2 complete - Workspace at specs/active/{slug}/"

Checkpoint 3: Intelligent Analysis

Use appropriate tool based on complexity:

• Simple: 10 structured thoughts
• Medium: Sequential thinking (15 thoughts)
• Complex: zen_planner or thinkdeep

For panels: Analyze Panel ABC interface, existing panels, template patterns For integrations: Analyze plugin protocol, middleware patterns, config inheritance

Document in workspace.

Output: "✓ Checkpoint 3 complete - Analysis using [tool]"

Checkpoint 4: Research (2000+ words)

Priority order:

1. Pattern Library: specs/guides/patterns/
2. Internal Guides: specs/guides/
3. Context7: Litestar/SQLAlchemy documentation
4. WebSearch: Best practices

For panel development:

mcp__context7__get-library-docs(
    context7CompatibleLibraryID="/litestar-org/litestar",
    topic="middleware",
    mode="code"
)

Verify: wc -w specs/active/{slug}/research/plan.md

Output: "✓ Checkpoint 4 complete - Research (2000+ words)"

Checkpoint 5: Write PRD (3200+ words)

Include:

• Intelligence context (complexity, similar features, patterns)
• Problem statement
• Acceptance criteria (specific, measurable)
• Technical approach with pattern references
• Testing strategy (90%+ coverage target)
• Files to create/modify

Template:

# PRD: {Feature Name}

## Metadata
- Slug: {slug}
- Complexity: {simple|medium|complex}
- Checkpoints: {6|8|10+}

## Intelligence Context
### Similar Implementations
- ...

### Patterns to Follow
- ...

## Problem Statement
...

## Acceptance Criteria
1. [ ] ...
2. [ ] ...

## Technical Approach
...

## Files to Create/Modify
- `src/debug_toolbar/...`
- `tests/unit/...`

## Testing Strategy
- Coverage target: 90%+
- Key test cases: ...

## Out of Scope
- ...

Verify: wc -w specs/active/{slug}/prd.md

Output: "✓ Checkpoint 5 complete - PRD (3200+ words)"

Checkpoint 6: Task Breakdown

Adapt to complexity level.

Simple (6 checkpoints):

1. Implement core functionality
2. Add tests
3. Update config if needed

Medium (8 checkpoints):

1. Create base class/module
2. Implement core logic
3. Add lifecycle hooks
4. Write unit tests
5. Write integration tests
6. Update documentation

Complex (10+ checkpoints):

• Full breakdown with dependencies

Output: "✓ Checkpoint 6 complete - Tasks adapted to complexity"

Checkpoint 7: Recovery Guide

Create specs/active/{slug}/RECOVERY.md with:

• Intelligence context for session resumption
• Current progress state
• How to continue from any checkpoint

Output: "✓ Checkpoint 7 complete - Recovery guide with intelligence context"

Checkpoint 8: Git Verification

git status --porcelain src/ | grep -v "^??"

Verify no source code was modified during PRD phase.

Output: "✓ Checkpoint 8 complete - No source code modified"

Final Summary

PRD Phase Complete ✓

Workspace: specs/active/{slug}/
Complexity: [simple|medium|complex]
Checkpoints: [6|8|10+] completed

Intelligence:
- ✓ Pattern library consulted
- ✓ Similar features analyzed
- ✓ Tool selection optimized

Next: Run `/implement {slug}`