Brand images on demand

Say what you want; the brand look comes free. In the command bar (typed or spoken):

• “generate an image of a hero banner for the launch page”
• “make me a brand image for the TV guide announcement”
• “image: a night skyline in our style”

The finished image lands on the canvas as an image pane — draggable, resizable, and attachable to an agent like any other image. A copy is saved into the workspace’s assets folder.

Template presets

Name a preset and the composition comes pre-designed (the brand still applies):

• “generate a hero image for the launch page” — wide 16:9, headline space
• “make an og card about the TV guide” — bold at thumbnail size
• “create a social square for the v2 release” — punchy 1:1
• “generate a logo mark of a fortress gate” — single centered glyph, 1:1
• “make a feature frame of the dashboard” — staged product scene, 4:3

Add your own: drop a JSON file ({"id","label","prompt","aspect"}, with a {{subject}} slot) into the app’s media-templates/ folder — it merges with the built-ins by id.

Agents can request media too

Your agents know the same trick. When an agent’s reply includes a directive alone on its own line, Mastery HQ renders it automatically once the run finishes:

• [[image: a hero banner for the leaderboard]]
• [[image logo-mark: a fortress glyph]] — any template preset works
• [[video: waves on a neon shore]] — an AI-generated clip (takes minutes)
• [[reel: neon skyline | ship faster]] — an AI clip finished with the brand title overlay
• [[bumper: ship faster]] — the 5-second brand title video

Ask an agent to “write the launch post and request a hero image for it” and the image arrives on the canvas alongside the post. Requests go through the exact same pipeline as your own commands — Brand Kit injection, the daily cap, the confirm card, and an Audit Log entry naming which agent asked. At most three requests fire per reply.

The brand is automatic

Every generation silently appends your Brand Kit (Settings → Brand): palette, typography feel, imagery style, tone, and the things to avoid. Tune the kit once and every image matches it — the tab’s “What generations receive” preview shows the exact text models get.

Costs stay visible and capped

• The success toast shows the actual billed cost of each image.
• Spending counts against the daily cap in Settings → Brand; when it’s reached, generation stops for the day.
• A generation estimated above your confirm threshold asks before running — an in-app confirm card showing the estimate and today’s spend. Left unanswered (say, a remote Hermes trigger while you’re away), it cancels itself after 90 seconds.

Setup

One key: add OPENROUTER_API_KEY to your .env (Settings → Agents → Open .env file). That single key reaches 30+ image models; without it, the command explains exactly what’s missing.