Browse local marketing assets with ak:content-hub

Use ak:content-hub to scan a project's assets/ directory and browse the result in a local visual gallery. The Hub can filter and preview assets, show brand context, rescan the library, and edit supported text files. Its manifest contains fields reserved for Cloudflare R2, but cloud sync is not active.

Choose ak:content-hub for a local asset library

Use ak:content-hub when

• Marketing assets already live under the current project's assets/ folder.
• You want visual search, type filters, previews, paths, and basic metadata.
• You want to compare assets with colors, typography, or voice extracted from an existing brand-guideline file.
• You accept a local HTTP server and can review any local file save.

Choose another workflow when

• You need content strategy or an editorial calendar. Use ak:content-marketing.
• You need cloud asset storage or R2 synchronization. The Hub only records R2 fields in .assets/manifest.json; its upload control is disabled.
• You need a full digital-asset-management system with permissions, approvals, versioning, or remote collaboration.

Prepare the project and runtime

Confirm Marketing Kit is installed for the current runtime and scope. Run the Skill from the project whose assets/ folder you intend to scan. The project must allow Node.js, a free local port, browser access, and writes to assets/, .assets/, and the temporary PID location. Create .assets/ if it does not exist; the shipped scanner writes the manifest but does not create its parent.

Brand context is read from the first available path among docs/brand-guidelines.md, docs/design-guidelines.md, and brand-guidelines.md. AI enhancement additionally requires the claude CLI to be available; gallery browsing does not.

Open the Hub

/ak:content-hub open Scan this project's assets, open the local gallery, and do not save or enhance any file until I approve it

/ak:content-hub browse Show the local asset library and brand context without changing asset files

$ak:content-hub search Find assets related to the launch campaign; keep the run local and read-only

The canonical action hint is open|browse|search. State your authority boundary in the request because the gallery itself includes save and AI-enhance actions.

Understand the stages

1. Start the local service. The bundled server chooses an available port from 3457 through 3500, records a PID, and can open /hub in a browser.
2. Scan the asset tree. The scanner reads supported files recursively, derives type, category, tags, size, and timestamps, then writes .assets/manifest.json while preserving existing R2 and user metadata.
3. Load brand context. The sidebar extracts a bounded set of colors, typography, and voice traits from the first supported guideline file.
4. Browse or filter. Search and type filters operate on the manifest; previews and copied paths refer to local files.
5. Edit only with approval. Supported text assets can be loaded and saved from the gallery. AI enhancement sends the current content and instruction through the local claude CLI before you decide whether to save it.

Keep file and provider effects explicit

Review sensitive data before scanning because paths and metadata enter the manifest and supported files can be served by the local HTTP process. Stop the service when finished. Do not expose its port beyond the local machine or treat the path check as a complete security boundary.

Verify outputs and evidence

A useful run should provide the local Hub URL, scanned asset count, manifest path, categories and formats found, brand-context source or a clear missing-file notice, and any file path changed. For a read-only browse request, the only project mutation should be the refreshed .assets/manifest.json.

If you approve an edit, inspect the before-and-after content yourself. A gallery preview, inferred tag, aiGenerated filename heuristic, or AI rewrite is not proof of asset quality, origin, rights, accuracy, or brand approval.

Troubleshoot and know the limits

The Hub handles a local asset library, not publication, remote sync, licensing, or source-of-truth governance. Continue with the Marketing Kit overview or review Runtime adapters before assuming equal behavior across runtimes.