create-meta-prompt

Create prompts optimized for Claude-to-Claude communication in multi-stage workflows. Outputs are structured with XML and metadata for efficient parsing by subsequent prompts.

Every execution produces a SUMMARY.md for quick human scanning. Each prompt gets its own folder in .prompts/ with output artifacts, enabling clear provenance and chain detection.

1. **Intake**: Determine purpose (Do/Plan/Research/Refine), gather requirements 2. **Chain detection**: Check `.prompts/*/` for existing research/plan files to reference 3. **Generate**: Create prompt using purpose-specific patterns 4. **Save**: Create folder `.prompts/{number}-{topic}-{purpose}/` 5. **Present**: Show decision tree for running 6. **Execute**: Run prompt(s) with dependency-aware execution 7. **Summarize**: Create SUMMARY.md for human scanning

<folder_structure>

.prompts/
├── 001-auth-research/
│   ├── completed/001-auth-research.md   # archived after run
│   ├── auth-research.md                 # full output (XML for Claude)
│   └── SUMMARY.md                       # executive summary (markdown for human)
├── 002-auth-plan/
│   ├── completed/002-auth-plan.md
│   ├── auth-plan.md
│   └── SUMMARY.md
└── 003-auth-implement/
    ├── completed/003-auth-implement.md
    └── SUMMARY.md

</folder_structure>

If no context provided, ask for purpose: Do / Plan / Research / Refine.

If context provided, infer purpose from keywords:

• implement, build, create, fix → Do
• plan, roadmap, strategy, phases → Plan
• research, understand, analyze, explore → Research
• refine, improve, iterate, update → Refine

Extract topic identifier (kebab-case, e.g. auth, stripe-payments). Ask 2-4 clarifying questions based on purpose gaps. Confirm before generating.

<prompt_structure> All generated prompts include:

1. Objective — what to accomplish and why
2. Context — referenced files (@), dynamic context (!)
3. Requirements — specific instructions
4. Output specification — save location and structure
5. Metadata (Research/Plan only) — XML tags required in output
6. SUMMARY.md requirement — all prompts must produce one
7. Success criteria — how to verify it worked

For Research and Plan outputs, require these XML tags:

• <confidence> — confidence level in findings
• <dependencies> — what's needed to proceed
• <open_questions> — what remains uncertain
• <assumptions> — what was assumed

SUMMARY.md must contain: one-liner, Key Findings, Decisions Needed, Blockers, Next Step. </prompt_structure>

Detect dependencies by scanning prompts for `@.prompts/{number}-{topic}/` references.

Sequential: chained prompts where each depends on previous output — execute in order, stop on failure.

Parallel: independent prompts with no dependencies — spawn all Task agents in a single message.

Mixed: build dependency layers, run parallel within each layer, sequential between layers.

After each completion: validate output exists, is non-empty, has required metadata tags and SUMMARY.md. Archive prompt to completed/ subfolder on success.

Show inline SUMMARY.md content in results so user sees findings without opening files.

<failure_handling> Sequential failure: stop chain, report completed/failed/not-started, offer retry or stop.

Parallel failure: continue others, collect all results, report failures with details, offer retry.

Missing dependencies: warn user, offer to create missing prompt first or continue anyway. </failure_handling>

<success_criteria>

• Purpose and topic identified during intake
• Chain detection performed, relevant files referenced
• Prompt generated with correct structure for purpose
• Folder created in .prompts/ with correct naming
• SUMMARY.md requirement and metadata requirements included
• Prompts executed in correct dependency order
• Output validated after each completion
• Successful prompts archived to completed/
• SUMMARY.md displayed inline in results </success_criteria>