agency-multi-brand-pack

Build a manifest listing each client slug with its templates, expand it into flat jobs with IDs like client-slug/asset-name, then run gen-ai batch run with a dry-run estimate first. The slash in job IDs automatically creates per-client output subfolders.

Attach each client's brand rules at the job level, not the batch level, so every prompt carries the correct brand.md constraints. Also namespace all outputs by client slug to avoid filename collisions between clients.

Use gen-ai batch resume with the output directory to continue from where the run stopped instead of restarting. This preserves completed jobs and avoids paying twice for finished assets.

Tag every job with the client slug in its ID, then parse results.json to group completed and failed jobs per client. Combine this with gen-ai pricing data to produce a per-client spend breakdown for finance.

Avoid it for single-client deep-production work, where a client-specific workflow or pitch-mockup skill fits better. Multi-brand batching is designed for portfolio parallelism, not one-off detailed creative.