Export Prompt Presets: Minimal CTA, External Intelligence

Organic LLM now includes a centralized export flow that lets users move selected context into external systems with almost no UI friction.

The design goal is simple:

  • keep the interface minimal,
  • keep user intent explicit,
  • keep output structure reliable.

This post documents how the feature works and why it is structured this way.

Why This Feature Exists

Users often need outside capabilities while staying grounded in their current workspace context:

  • script polishing,
  • architecture summarization,
  • career profile extraction,
  • code analysis in Cursor.

The previous pattern required manual prompt writing and context reformatting. The new export flow reduces this to one action.

UX Principle: Minimal by Default

The user should primarily see a single clear CTA, such as:

  • Request Script from ChatGPT
  • Analyze code with Cursor

We avoid exposing too many intermediate controls in the main interaction path. Prompt complexity is handled by Organic LLM behind the scenes.

Architecture Summary

The feature is split into three layers:

  1. Preset registry

    • canonical prompt intents and output contracts
    • file: lib/export/prompts.ts

  2. Reusable modal

    • centralized text display + copy + export actions
    • file: components/design-system/TextCopyModal.tsx

  3. Prompt generation route

    • Organic LLM generates/refines the final external prompt
    • file: app/api/ai/export-prompt/route.ts

First adopter:

  • app/sandbox/prototypes/glass-fonts/_components/glass-fonts-lab.tsx

Preset-Driven Prompt Contracts

Each preset defines:

  • target app (chatgpt or cursor),
  • user-facing button label,
  • intent question,
  • required response format,
  • optional context placeholder.

This keeps exports predictable and easier to parse when brought back into Organic LLM.

Organic LLM as Prompt Author

Instead of directly sending template text, the feature can route through Organic LLM to generate a refined external prompt:

  • source context from the modal,
  • selected preset contract,
  • optional user context,
  • target-app-specific framing.

If generation fails, deterministic fallback builders still produce a valid prompt/instruction package.

Target Behaviors

ChatGPT

  • Generates final prompt text.
  • Copies prompt to clipboard.
  • Opens ChatGPT with prefilled query text.

Cursor

  • Generates structured instruction package.
  • Copies instruction to clipboard for immediate paste into Cursor.

Current Cursor integration is clipboard-first by design. Deep-link/agent handoff can be added later without changing preset structure.

Design Notes

  • Modal background uses heavy blur with subtle darkening.
  • CTA labels are explicit and task-oriented.
  • Prompt structure is centralized for maintainability.
  • The system is extensible without per-page hardcoding.

What Comes Next

  • Additional preset packs by workflow type.
  • Optional return-channel schema for importing outputs back into Organic LLM.
  • Richer app-specific integrations while preserving the same preset contract model.