Artifact Review UX Roadmap

implementation plan 2026-03-15 3 min read 557 words
Artifact Preview

Artifact Review UX Roadmap

Executive Summary

This roadmap defines UX improvements for reviewing agent-generated artifacts across Discord mobile, Discord desktop, and Mission Control on the web. The core problem is that raw markdown is a poor delivery format on mobile, so the system must package outputs for review rather than shipping plaintext blobs. The phased build covers rendered HTML previews, PDF exports, and a dedicated artifact detail reader in Mission Control.

Goal

Make agent outputs easy to review from:

The real problem

Raw markdown is a lousy delivery format in Discord. On mobile it is worse. If the output matters, the system should package it for review instead of making Pete wrestle with plaintext.

Design rules

  1. Main channel gets summary, not the whole document.
  2. Every substantial artifact gets at least one rendered preview path.
  3. Mobile review should work even if Mission Control links are inconvenient.
  4. Laptop review should feel like a document reader, not a note field.

For markdown-heavy artifacts, generate:

Discord delivery modes

Mode 1: summary + image, default

Use when Pete mainly needs to scan and decide.

Post:

Mode 2: summary + PDF

Use when the document is dense and likely needs full reading on mobile.

Post:

Mode 3: summary + thread chunks

Use only when Pete explicitly wants the text in Discord.

Rules:

Clever options worth testing

Option A: card image generated from markdown headings and key bullets

Pros:

Cons:

Option B: rendered HTML served by Mission Control

Pros:

Cons:

Option C: PDF snapshot from HTML

Pros:

Cons:

Pros:

Cons:

Option E: stitched section images for ultra-mobile review

Generate one image per major section or first two pages.

Pros:

Cons:

Mission Control artifact reader requirements

Best first implementation

  1. HTML preview generation
  2. artifact detail page
  3. summary image generation
  4. PDF export
  5. Discord delivery helper

Success criteria

The UX is good enough when Pete can: