/prompt-creator
Expert prompt engineering for creating effective prompts for Claude, GPT, and other LLMs.
The /prompt-creator skill creates prompts optimized for Claude-to-Claude communication in multi-stage workflows. Each prompt gets its own folder in .prompts/ with structured outputs and human-readable summaries.
Prompt Purposes
| Purpose | Description | Use When |
|---|---|---|
| Do | Execute a task, produce an artifact | implement, build, create, fix, add, refactor |
| Plan | Create an approach, roadmap, or strategy | plan, roadmap, approach, strategy, decide |
| Research | Gather information or understand something | research, understand, learn, gather, analyze |
| Refine | Improve an existing research or plan output | refine, improve, deepen, expand, iterate |
Workflow
- Intake: Determine purpose, gather requirements via adaptive questions
- Chain detection: Scan for existing research/plan files to reference
- Generate: Create prompt using purpose-specific patterns
- Save: Create folder in
.prompts/{number}-{topic}-{purpose}/ - Present: Show decision tree for running
- Execute: Run prompt(s) with dependency-aware execution
- Summarize: Validate output and create SUMMARY.md
Folder Structure
| Path | Contents |
|---|---|
.prompts/{number}-{topic}-{purpose}/ | Root folder for each prompt |
completed/ | Archived prompts after execution |
{number}-{topic}-{purpose}.md | The prompt (moved to completed/ after run) |
{topic}-{purpose}.md | Full XML-structured output for Claude |
SUMMARY.md | Human-readable executive summary |
Example:
.prompts/
├── 001-auth-research/
│ ├── completed/001-auth-research.md # Archived prompt
│ ├── auth-research.md # Full output (XML)
│ └── SUMMARY.md # Executive summary
├── 002-auth-plan/
│ ├── completed/002-auth-plan.md
│ ├── auth-plan.md
│ └── SUMMARY.md
└── 003-auth-implement/
├── completed/003-auth-implement.md
└── SUMMARY.md # Implementation creates code elsewhere
Execution Modes
| Mode | When Used | Behavior |
|---|---|---|
| Single | One prompt | Straightforward execution |
| Sequential | Chain with dependencies | Execute in order, stop on failure |
| Parallel | Independent prompts | Run all at once, continue despite failures |
| Mixed | Complex DAG | Layer-based execution (parallel within layers, sequential between) |
Dependencies are auto-detected from @ references in prompts or inferred from purpose.
Output Validation
After execution, prompts are validated for:
- Output file exists and has content (> 100 chars)
- Metadata present (Research/Plan only):
<confidence>,<dependencies>,<open_questions>,<assumptions> - SUMMARY.md exists with required sections
- One-liner is substantive (not generic like "Research completed")
SUMMARY.md Structure
| Section | Description |
|---|---|
| One-liner | Substantive description of outcome |
| Version | v1 or iteration info |
| Key Findings | Actionable takeaways |
| Files Created | (Do prompts only) |
| Decisions Needed | What requires user input |
| Blockers | External impediments |
| Next Step | Concrete forward action |
When to Use
Use /create-meta-prompts when:
- Building multi-stage workflows (research → plan → implement)
- Creating prompts that produce outputs for other prompts
- Needing clear provenance and chain detection
- Want structured outputs with validation
Reference Guides
The skill uses purpose-specific patterns from:
references/do-patterns.md- Execution promptsreferences/plan-patterns.md- Planning promptsreferences/research-patterns.md- Research promptsreferences/refine-patterns.md- Iteration promptsreferences/summary-template.md- SUMMARY.md structurereferences/metadata-guidelines.md- XML metadata requirements