
Oracle Prompt Library
A centralized repository to streamline prompt creation, reuse, and governance across Oracle’s AI tools and developer experiences.
Context & Opportunity
With the rapid adoption of LLMs across Oracle’s suite of products (e.g., SQL Developer, APEX, Code Assist), teams were independently hardcoding prompts, resulting in:
Redundancy (tonality, syntax, formatting)
Inconsistent user experiences
No shared taxonomy or central governance
There was no scalable way to share or version prompts across teams.
Goal
Design a centralized internal library to:
Standardize prompt formatting across tools
Enable teams to author, reuse, and fork prompts
Integrate with Oracle’s LLM interfaces (REST API, SQLDev, APEX, Gen AI)
My Role
Product Designer
I led the experience strategy, information architecture, and component-level UX, collaborating closely with AI PMs, staff engineers, and developer advocates. I was the sole designer on the project.
Discovery
We mapped friction points with stakeholders from SQLDev, APEX, ORDS, and MCP. Key insights:
Hardcoded prompts made iteration painful
No prompt-to-product traceability
Repeated structures across teams (e.g. “task,” “tone,” “language style”)
Demand for LLM-specific variations and version control
Core Features
Designed and prototyped the following:
Prompt Library Dashboard
Filter by model, category, popularity
Search and sort by relevance or trending
Clear UX for 1-click “View,” “Save,” and “Fork”
Prompt View Page
Markdown output
Revision history and model compatibility
Engagement metrics (likes, forks, API calls)
Prompt Editor
Draft mode with autosave
Sectional editor with Markdown and preview
Selectable templates or build-your-own
Support for versioning and publishing rules
Bookmarks & Stats
Save for reuse
View prompt performance (e.g. API usage, edits, comments)
Design Highlights
Applied Oracle’s internal design system for interface cohesion
Used semantic color-coding and layout grids for high scanability
Defined metadata architecture: model compatibility, version, team ownership
Built for scale — supports 3,000+ prompts, REST API access, and multiple LLMs (ChatGPT, Claude, Copilot, Ollama)

