Marketing Lab
One brief, specialised producers, a shared quality loop.
The business problem
Small teams need consistent brand, print, web, social and video assets without coordinating a separate production process for each format.
My contribution
My role covered workflow definition, architecture, direction of AI-assisted implementation, service integration, testing and deployment. Platform and library authorship belongs to the third parties credited below; additional collaborators are not identified in the public project notes.
Claude Code, Canva MCP and Remotion are existing tools. Contribution centres on orchestration and quality rules.
Architecture & boundaries
A READY marker triggers nine specialised Claude Code agents around a shared brand.json. Producers write to isolated folders; a QA guardian can re-trigger only the failed producer.
- 01BriefShared brand.json
- 02ProduceIsolated specialist outputs
- 03ReviewQA guardian re-triggers a failed producer
Labelled architecture overview, reconstructed from project documentation. This is not a production trace.
Failure paths & design decisions
One bad output should not trigger a complete re-run.
Separate producer workspaces and re-trigger the failed producer.
Evidence boundary: Described in the project notes; no public failure/recovery trace is available.
Delivery record
- Status
- A documented test pack, with broader ongoing production use unverified.
- Users & scope
- Brand, print, web, social, video and copy production around a shared brief.
- Dates & usage
- Test date and repeat-run results are not published.
Results & measurement gaps
A documented test pack produced 22 deliverables in 12 minutes. That is a single test result, rather than a promise of production throughput.
22 deliverables in 12 minutes is a single documented test result, not a sustained throughput or cost benchmark.