How DCP Flow works
Explains the DCP Flow operating model, its pragmatic Scrum adaptation, core artifacts, Golden Flows, evidence, and human acceptance.
Explains the DCP Flow operating model, its pragmatic Scrum adaptation, core artifacts, Golden Flows, evidence, and human acceptance.
DCP Flow operates as an iterative work system combining human decisions, explicit artifacts, AI-assisted execution, and verifiable evidence. It borrows from Scrum while reducing ceremony and adding controls for agent-assisted engineering.
A pragmatic operating model
Sprints target verifiable outcomes rather than task volume. The team defines an objective, limits scope, executes with shared context, and reviews evidence before accepting the result.
- Iteration and backlog with less ceremony.
- Sprint Goal as a verifiable outcome.
- Human Gate for material decisions and acceptance.
- Retrospective and learning close the loop.
Artifacts that connect intent and execution
Constitution, Standards, Blueprints, Skills, ADRs, Golden Flows, Quality Gates, and Project Truth serve different purposes. A conversation or AI response is never treated as project truth by itself.
- Golden Flow: critical end-to-end journey.
- Evidence: observed or executed proof.
- Quality Gate: minimum evidence before progress.
- Repository Truth: durable auditable state.
Scrum adapted to DCP context
DCP Flow keeps iteration, backlog, prioritization, review, and continuous improvement from Scrum, while reducing ceremonies and rigid roles when they add little value. A sprint is primarily a contract for a verifiable outcome rather than only a time box.
- Incremental work.
- Explicit backlog.
- Sprint as verifiable outcome.
- Less ceremony when risk does not justify it.
The complete operating cycle
The methodology connects outcome, discovery, ADOPT/CONFIGURE/INTEGRATE/EXTEND/BUILD decision, architecture, Blueprint and Standards, Golden Flow, execution, QA, evidence, human review, release, and learning. It preserves continuity between the original decision and the verifiable result.
- Outcome → Discovery → Decision.
- Architecture → Sprint → Execution.
- QA → Evidence → Human Review.
- Release → Support → Learning.
AI-native does not mean AI-autonomous
Kai and Execution Agents accelerate analysis and implementation without removing human authority. Material decisions, acceptance, risk, and authorization remain governed. Git, PRs, and evidence preserve durable truth beyond an agent conversation.
- Kai applies the methodology.
- Execution Agents perform bounded work.
- Human Acceptance preserves responsibility.
- Git/Evidence preserve traceability.
Contextual glossary＋
Critical end-to-end journey that must work and produce verifiable evidence.
Checkpoint where evidence determines whether work may progress.
Proof of an observed or executed result; not an agent assertion.
Point where a material decision requires human acceptance or authorization.