Prototype for a coding-agent-assisted data product builder using DPH-style context:
- asset discovery
- lineage reasoning
- contract/template inheritance
- policy recommendation
- draft build bundle generation
Given a request like: "Create a new Customer 360 UK data product from existing GCP assets"
the prototype should:
- discover candidate assets
- inspect lineage
- select a contract template
- recommend data protection rules
- produce:
- output/build_bundle.json
- output/draft_contract.yaml
- output/policy_recommendations.json
- output/prototype_plan.md
- Start from checked-in fixtures, not live platform calls
- Treat contracts and protection rules as governed inputs
- Do not directly mutate GCP services
- Produce proposals and drafts first
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
python scripts/build_bundle.py
python scripts/validate_contract.pyAsk Codex to:
- read AGENTS.md
- read docs/tool-map.md
- use the dph-builder skill
- generate the build bundle from fixtures
- keep policy actions in recommendation form