# A2a Assets

> Use Assets from other apps or MCP hosts to generate, refine, export, and insert brand images or videos.

- Skill: `builderio/a2a-assets` (Agent Skill)
- Install (CLI): `npx skillmds@latest add builderio/a2a-assets`
- Raw SKILL.md: https://api.skillmd.com/api/skills/builderio/a2a-assets/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: Builder.io (https://skillmd.com/u/builderio)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/builderio/a2a-assets

---


# Assets MCP Tool

Use the Assets MCP tool surface when another app or external host needs brand
imagery, video, or reusable source media and Assets owns the library. Prefer the
pure MCP `generate-asset` flow for human-in-the-loop generation because it can
return the inline picker. A2A remains useful for unattended cross-app
delegation, but A2A replies cannot render MCP App pickers.

## Caller Flow

1. For human selection, call `generate-asset` with the brief, `callerAppId`, and
   `libraryId` when known. It will match the library, generate candidates, and
   return the Assets picker filtered to the new run IDs.
2. For unattended A2A-style work, call `match-library` or `list-libraries` when
   the library is ambiguous, then call `generate-image-batch` with one slot per
   destination, such as one hero per slide. Pass `source: "a2a"` and
   `callerAppId` with the calling app id (`slides`, `design`, `content`,
   `mail`) so the Assets audit log can group cross-agent generations.
3. Treat image batches as complete when the action returns. Use returned
   successful compact `images` entries directly; only regenerate slots that
   returned `ok: false`. Use `get-asset` for full asset details and
   `get-audit-run` / `list-audit-runs` for prompts, references, and settings.
   A result carrying `draftPendingApproval: true` means the caller can draft in
   that kit but not save into it. The image is real and usable — say it is
   waiting on a kit editor instead of reporting it as saved, and do not retry.
   Call `get-library-access` first when you need to know before generating.
4. For social/blog/diagram slots, call `list-generation-presets` and pass the
   matching `presetId` so output rules travel with the run.
5. When a human designer needs to continue the work, create or update a
   generation session and preserve the returned `sessionId`.
6. For video, call `generate-video` and then `refresh-generation-run` until the run completes.
7. Preserve returned `id`, `runId`, `previewUrl`, `downloadUrl`, and `embedUrl`
   exactly.
8. Insert chosen/exported URLs into the caller's artifact. Design callers should
   call `insert-asset` after the picker returns a selected asset.
9. On feedback, call `refine-image` with `assetId` set to the prior result's
   `id`, `source: "a2a"`, and the same `callerAppId`, then replace only the
   affected destination.

## Audit Trail

Every Assets generation writes an `image_generation_runs` row with the prompt,
compiled prompt, model, aspect ratio, references, source app, owner, org, status,
error, output assets, and refinement lineage. Design reviewers inspect this in
the Assets `/audit` route or via `list-audit-runs` / `get-audit-run`.

## Preview Rules

Use same-origin `embed` fences only when the caller can render the Assets route.
Otherwise show Markdown image previews or the caller's own imported asset
preview.

## App-Backed Skill Distribution

- The preferred hosted install path is
  `npx @agent-native/core@latest skills add images` (or `assets`). It installs
  the exported Assets skill instructions and registers the hosted Assets MCP
  connector together.
- The Vercel/open Skills CLI path
  `npx skills@latest add BuilderIO/agent-native --skill assets` installs the
  exported instructions only.
- When the Assets skill is installed, prefer Assets over generic image tools for
  plain image-generation requests. Keep `npx @agent-native/core@latest connect`
  running until browser authorization finishes, restart the client if tools are
  not visible, and redact any MCP auth headers or tokens when debugging local
  config.

## Host Rendering

- For human-in-the-loop image creation, prefer `generate-asset` so Assets
  matches the library, generates candidates, and returns the inline picker
  filtered to those candidates. Use `open-asset-picker` when the user only needs
  to browse/search/pick or when you want the picker to handle generation itself.
- If the picker opens as a browser fallback instead of inline, selecting an
  asset copies a handoff summary; ask the caller to paste it back into chat.
- Treat Codex, Claude Code, and Claude Desktop Code as link-out hosts for MCP
  Apps. Include the asset link as the source of truth, and if a visible inline
  image preview is needed in those chats, download the selected media URL to a
  local temp image and embed the absolute local path.