Skip to content

Persona 1 — Human Domain Expert

Type: Human Accountable for: Correctness. If a contract is wrong, every downstream artifact is wrong.

What this persona authors

  • contracts/slas/ — SLA targets negotiated with stakeholders
  • contracts/domain-invariants/ — business rules the domain must never violate
  • contracts/event-schemas/ — canonical JSON Schemas for every domain event
  • tooling/service-manifest.yaml — authoritative list of services, owners, resources
  • docs/adrs/ — decision context and rationale stubs (the why, not the full MADR)
  • docs/ddd/bounded-contexts.md, ubiquitous-language.md — context boundaries and shared language
  • ai-agents/tasks/ — task specifications that instruct every other persona

What this persona does not author

Code, infrastructure manifests, observability artifacts, full ADRs, runbooks, migration docs. Those are produced by agents from the contracts this persona writes.

Review responsibility

This persona performs the final review gate after each agent phase. The review question is always the same: does this artifact correctly express the contract? Not: is this good code?

Guardrails

CODEOWNERS maps contracts/ to @{project}/senior-engineers. No agent-initiated PR may modify a contract file.

Contract Challenge Review

Review incoming Contract Challenge Reports (CCRs) in ai-agents/reviews/ccr-*.md. Route each to the relevant product or business owner with the cost signal. Record the outcome directly in the CCR file:

  • Relax: amend the contract file in contracts/ and re-enter Phase 0 for the affected contract.
  • Hold: write an ADR in docs/adrs/ documenting why the contract stands at this cost. Resume the originating phase once the ADR is written and reviewed.

Neither outcome is undocumented. Every CCR resolution produces a written artifact.