A structured library of system prompts, chained workflows (multi-agent pipelines), and evaluation criteria for AI content generation.
This repository translates complex content requirements—specifically for UX writing, technical documentation, and fintech compliance—into reliable, reproducible LLM parameters.
This library is organized into specialized workspaces based on the cognitive task required of the LLM:
Large language models hallucinate or lose nuance when asked to write long-form content in a single prompt. This directory contains a chained prompt pipeline that breaks generation into discrete steps:
Step 1: Core generation and merging context.Step 2-3: Injecting unique value, depth, and specific examples.Step 4: Structuring metadata (Charts, Alt-texts).Step 5: Linguistic proofreading and strict compliance/security checks.
Frameworks for generating and refining user interface copy at scale, maintaining tone consistency and systemic logic.
- Error message bulk generation.
- Product copy revision and glossary alignment.
- Transactional email logic.
Prompts designed not for drafting, but for synthesis, reasoning, and workflow management.
- UX research synthesis and Slack summarization.
- Document outlining and presentation structuring.
- Stakeholder alignment and feedback analysis.
Built by Himawan | Designed for GPT-4, Claude 3, and Gemini.