create-meta-prompts

Prompt chaining orchestrates sequential AI stages by automatically generating and linking prompts for research, planning, and execution. Each stage produces structured outputs with XML metadata (confidence, dependencies, open questions) that feed directly into the next stage, eliminating manual context switching and ensuring logical workflow progression.

Meta-prompting creates optimized prompts designed for Claude-to-Claude pipelines, automating the generation of prompts that produce structured, machine-readable outputs. Use it when you need to chain multiple AI analysis or planning stages together with preserved dependencies and clear provenance tracking across the workflow.

Structured outputs embed XML metadata blocks containing confidence levels, dependency references, open questions, and assumptions. Meta-prompting automates the creation of prompts that enforce this metadata format, ensuring each stage's output is machine-parseable and ready for downstream consumption by subsequent AI stages.

Automated dependency detection identifies and references outputs from prior workflow stages without manual linking. The system tracks which stages depend on others, generates per-prompt folders under .prompts with SUMMARY.md files, and ensures dependency-aware execution order across your multi-stage workflow.

Manual prompt writing requires crafting each stage separately and managing context transfer by hand. Meta-prompting automates prompt generation, dependency linking, and metadata embedding across all stages, reducing errors and ensuring consistency in structured outputs, XML formatting, and workflow coherence.

Claude Code provides the execution environment for orchestrating chained prompts, but meta-prompting itself is the technique for generating prompts that work together. The Skill applies to any workflow where you're automating multi-stage prompt creation for research, planning, or implementation tasks with structured metadata.